,
Tanguy Barthelemy [aut, cre, art],
Anna Smyk [aut]| Title: | Utility Functions around 'JDemetra+ 3.0' |
|---|---|
| Description: | R Interface to 'JDemetra+ 3.x' (<https://github.com/jdemetra>) time series analysis software. It provides functions allowing to model time series (create outlier regressors, user-defined calendar regressors, UCARIMA models...), to test the presence of trading days or seasonal effects and also to set specifications in pre-adjustment and benchmarking when using rjd3x13 or rjd3tramoseats. |
| Authors: | Jean Palate [aut],
Alain Quartier-la-Tente [aut] ,
Tanguy Barthelemy [aut, cre, art],
Anna Smyk [aut] |
| Maintainer: | Tanguy Barthelemy <[email protected]> |
| License: | file LICENSE |
| Version: | 3.3.0 |
| Built: | 2024-10-28 16:26:12 UTC |
| Source: | https://github.com/rjdverse/rjd3toolkit |
Information on the (log-)likelihood
.likelihood( nobs, neffectiveobs = NA, nparams = 0, ll, adjustedll = NA, aic, aicc, bic, bicc, ssq ).likelihood( nobs, neffectiveobs = NA, nparams = 0, ll, adjustedll = NA, aic, aicc, bic, bicc, ssq )
nobs |
Number of observation |
neffectiveobs |
Number of effective observations. NA if it is the same as nobs. |
nparams |
Number of hyper-parameters |
ll |
Log-likelihood |
adjustedll |
Adjusted log-likelihood when the series has been transformed |
aic |
AIC |
aicc |
AICC |
bic |
BIC |
bicc |
BIC corrected for the length |
ssq |
Sum of the squared residuals |
These functions are used in all JDemetra+ 3.0 packages to easily interact between R and Java objects.
.r2jd_tsdata(s) .r2jd_tsdomain(period, startYear, startPeriod, length) .jd2r_tsdata(s) .jd2r_mts(s) .jd2r_lts(s) .jd2r_matrix(s) .r2jd_matrix(s) .jdomain(period, start, end) .enum_sextract(type, p) .enum_sof(type, code) .enum_extract(type, p) .enum_of(type, code, prefix) .r2p_parameter(r) .p2r_parameter(p) .r2p_parameters(r) .r2p_lparameters(r) .p2r_parameters(p) .p2r_parameters_rslt(p) .p2r_parameters_rsltx(p) .p2r_test(p) .p2r_matrix(p) .p2r_tsdata(p) .r2p_tsdata(r) .p2r_parameters_estimation(p) .p2r_likelihood(p) .p2r_date(p) .r2p_date(s) .p2r_span(span) .r2p_span(rspan) .p2r_arima(p) .p2r_ucarima(p) .p2r_spec_sarima(spec) .r2p_spec_sarima(r) .p2r_outliers(p) .r2p_outliers(r) .p2r_sequences(p) .r2p_sequences(r) .p2r_iv(p) .r2p_iv(r) .p2r_ivs(p) .r2p_ivs(r) .p2r_ramps(p) .r2p_ramps(r) .p2r_uservars(p) .r2p_uservars(r) .p2r_variables(p) .p2r_sa_decomposition(p, full = FALSE) .p2r_sa_diagnostics(p) .p2r_spec_benchmarking(p) .r2p_spec_benchmarking(r) .r2jd_sarima(model) .jd2r_ucarima(jucm) .p2jd_calendar(pcalendar) .r2p_calendar(r) .proc_numeric(rslt, name) .proc_vector(rslt, name) .proc_int(rslt, name) .proc_bool(rslt, name) .proc_ts(rslt, name) .proc_str(rslt, name) .proc_desc(rslt, name) .proc_test(rslt, name) .proc_parameter(rslt, name) .proc_parameters(rslt, name) .proc_matrix(rslt, name) .proc_data(rslt, name) .proc_dictionary(name) .proc_dictionary2(jobj) .proc_likelihood(jrslt, prefix) .r2p_moniker(r) .p2r_moniker(p) .r2p_datasupplier(name, r) .p2r_metadata(p) .r2p_metadata(r, type) .p2r_ts(p) .r2p_ts(r) .p2r_tscollection(p) .r2p_tscollection(r) .r2jd_ts(s) .jd2r_ts(js) .r2jd_tscollection(s) .jd2r_tscollection(js) .p2r_datasupplier(p) .r2p_datasuppliers(r) .p2r_datasuppliers(p) .p2jd_variables(p) .jd2p_variables(jd) .jd2r_variables(jcals) .r2jd_variables(r) .p2r_context(p) .r2p_context(r) .p2jd_context(p) .jd2p_context(jd) .jd2r_modellingcontext(jcontext) .r2jd_modellingcontext(r) .p2r_calendars(p) .r2p_calendars(r) .p2jd_calendars(p) .jd2p_calendars(jd) .jd2r_calendars(jcals) .r2jd_calendars(r) .jd3_object(jobjRef, subclasses = NULL, result = FALSE) .p2r_regarima_rslts(p) .r2jd_tmp_ts(s, name) .r2jd_make_ts(source, id, type = "All") .r2jd_make_tscollection(source, id, type = "All") DATE_MIN DATE_MAX.r2jd_tsdata(s) .r2jd_tsdomain(period, startYear, startPeriod, length) .jd2r_tsdata(s) .jd2r_mts(s) .jd2r_lts(s) .jd2r_matrix(s) .r2jd_matrix(s) .jdomain(period, start, end) .enum_sextract(type, p) .enum_sof(type, code) .enum_extract(type, p) .enum_of(type, code, prefix) .r2p_parameter(r) .p2r_parameter(p) .r2p_parameters(r) .r2p_lparameters(r) .p2r_parameters(p) .p2r_parameters_rslt(p) .p2r_parameters_rsltx(p) .p2r_test(p) .p2r_matrix(p) .p2r_tsdata(p) .r2p_tsdata(r) .p2r_parameters_estimation(p) .p2r_likelihood(p) .p2r_date(p) .r2p_date(s) .p2r_span(span) .r2p_span(rspan) .p2r_arima(p) .p2r_ucarima(p) .p2r_spec_sarima(spec) .r2p_spec_sarima(r) .p2r_outliers(p) .r2p_outliers(r) .p2r_sequences(p) .r2p_sequences(r) .p2r_iv(p) .r2p_iv(r) .p2r_ivs(p) .r2p_ivs(r) .p2r_ramps(p) .r2p_ramps(r) .p2r_uservars(p) .r2p_uservars(r) .p2r_variables(p) .p2r_sa_decomposition(p, full = FALSE) .p2r_sa_diagnostics(p) .p2r_spec_benchmarking(p) .r2p_spec_benchmarking(r) .r2jd_sarima(model) .jd2r_ucarima(jucm) .p2jd_calendar(pcalendar) .r2p_calendar(r) .proc_numeric(rslt, name) .proc_vector(rslt, name) .proc_int(rslt, name) .proc_bool(rslt, name) .proc_ts(rslt, name) .proc_str(rslt, name) .proc_desc(rslt, name) .proc_test(rslt, name) .proc_parameter(rslt, name) .proc_parameters(rslt, name) .proc_matrix(rslt, name) .proc_data(rslt, name) .proc_dictionary(name) .proc_dictionary2(jobj) .proc_likelihood(jrslt, prefix) .r2p_moniker(r) .p2r_moniker(p) .r2p_datasupplier(name, r) .p2r_metadata(p) .r2p_metadata(r, type) .p2r_ts(p) .r2p_ts(r) .p2r_tscollection(p) .r2p_tscollection(r) .r2jd_ts(s) .jd2r_ts(js) .r2jd_tscollection(s) .jd2r_tscollection(js) .p2r_datasupplier(p) .r2p_datasuppliers(r) .p2r_datasuppliers(p) .p2jd_variables(p) .jd2p_variables(jd) .jd2r_variables(jcals) .r2jd_variables(r) .p2r_context(p) .r2p_context(r) .p2jd_context(p) .jd2p_context(jd) .jd2r_modellingcontext(jcontext) .r2jd_modellingcontext(r) .p2r_calendars(p) .r2p_calendars(r) .p2jd_calendars(p) .jd2p_calendars(jd) .jd2r_calendars(jcals) .r2jd_calendars(r) .jd3_object(jobjRef, subclasses = NULL, result = FALSE) .p2r_regarima_rslts(p) .r2jd_tmp_ts(s, name) .r2jd_make_ts(source, id, type = "All") .r2jd_make_tscollection(source, id, type = "All") DATE_MIN DATE_MAX
s |
Time series |
startYear |
Initial year in the time domain |
startPeriod |
Initial period in the time domain(1 for the first period) |
length |
Length |
p, r, spec, jucm, start, end, name, period, type, code, prefix, span, rspan, full, rslt, jd, jcontext, jobjRef, jcals, subclasses, result, pcalendar
|
parameters. |
model |
Model |
jobj |
Java object |
jrslt |
Reference to a Java object |
js |
Java time series |
source |
Source of the time series information |
id |
Identifier of the time series information (source-dependent) |
An object of class Message of length 3.
An object of class Message of length 3.
Title
.tsmoniker(source, id).tsmoniker(source, id)
source |
Source of the time series. |
id |
Id of the time series. |
Retail trade statistics in Australia
ABSABS
An object of class data.frame with 425 rows and 22 columns.
ABS
Generic function to add outliers or Ramp regressors (add_outlier() and
add_ramp()) to a specification or to remove them
(remove_outlier() and remove_ramp()).
add_outlier(x, type, date, name = sprintf("%s (%s)", type, date), coef = 0) remove_outlier(x, type = NULL, date = NULL, name = NULL) add_ramp(x, start, end, name = sprintf("rp.%s - %s", start, end), coef = 0) remove_ramp(x, start = NULL, end = NULL, name = NULL)add_outlier(x, type, date, name = sprintf("%s (%s)", type, date), coef = 0) remove_outlier(x, type = NULL, date = NULL, name = NULL) add_ramp(x, start, end, name = sprintf("rp.%s - %s", start, end), coef = 0) remove_ramp(x, start = NULL, end = NULL, name = NULL)
x |
the specification to customize, must be a "SPEC" class object (see details). |
type, date
|
type and date of the outliers. Possible |
name |
the name of the variable (to format print). |
coef |
the coefficient if needs to be fixed. If equal to 0 the outliers/ramps coefficients are estimated. |
start, end
|
dates of the ramp regressor. |
x specification parameter must be a JD3_X13_SPEC" class object
generated with rjd3x13::x13_spec() (or "JD3_REGARIMA_SPEC" generated
with rjd3x13::spec_regarima() or "JD3_TRAMOSEATS_SPEC" generated with
rjd3tramoseats::spec_tramoseats() or "JD3_TRAMO_SPEC" generated with
rjd3tramoseats::spec_tramo()). If a Seasonal adjustment process is
performed, each type of Outlier will be allocated to a pre-defined component
after the decomposition: "AO" and "TC" to the irregular, "LS" and Ramps to
the trend.
More information on outliers and other auxiliary variables in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/
add_usrdefvar, intervention_variable
# init_spec <- rjd3x13::x13_spec("RSA5c") # new_spec<-rjd3toolkit::add_outlier(init_spec, type="AO", date="2012-01-01") # ramp on year 2012 # new_spec<-rjd3toolkit::add_ramp(init_spec,start="2012-01-01",end="2012-12-01")# init_spec <- rjd3x13::x13_spec("RSA5c") # new_spec<-rjd3toolkit::add_outlier(init_spec, type="AO", date="2012-01-01") # ramp on year 2012 # new_spec<-rjd3toolkit::add_ramp(init_spec,start="2012-01-01",end="2012-12-01")
Function allowing to add any user-defined regressor to a specification and
allocate its effect to a selected component, excepted to the calendar component.
To add user-defined calendar regressors, set_tradingdays. Once added to
a specification, the external regressor(s) will also have to be added to a modelling context
before being used in an estimation process. see modelling_context and example.
add_usrdefvar( x, group = "r", name, label = paste0(group, ".", name), lag = 0, coef = NULL, regeffect = c("Undefined", "Trend", "Seasonal", "Irregular", "Series", "SeasonallyAdjusted") )add_usrdefvar( x, group = "r", name, label = paste0(group, ".", name), lag = 0, coef = NULL, regeffect = c("Undefined", "Trend", "Seasonal", "Irregular", "Series", "SeasonallyAdjusted") )
x |
the specification to customize, must be a "SPEC" class object (see details). |
group, name
|
the name of the regressor in the format |
label |
the label of the variable to be displayed when printing specification or results. By default equals to |
lag |
integer defining if the user-defined variable should be lagged.
By default ( |
coef |
the coefficient, if needs to be fixed. |
regeffect |
component to which the effect of the user-defined variable will be assigned.
By default ( |
x specification parameter must be a JD3_X13_SPEC" class object generated with rjd3x13::x13_spec()
(or "JD3_REGARIMA_SPEC" generated with rjd3x13::spec_regarima() or "JD3_TRAMOSEATS_SPEC"
generated with rjd3tramoseats::spec_tramoseats() or "JD3_TRAMO_SPEC" generated with
rjd3tramoseats::spec_tramo()).
Components to which the effect of the regressor can be allocated:
"Undefined" : the effect of the regressor is assigned to an additional component, the variable is used to improve the pre-processing step, but is not removed from the series for the decomposition.
"Trend": after the decomposition the effect is allocated to the trend component, like a Level-Shift
"Irregular": after the decomposition the effect is allocated to the irregular component, like an Additive-outlier.
"Seasonal": after the decomposition the effect is allocated to the seasonal component, like a Seasonal-outlier
"Series": after the decomposition the effect is allocated to
the raw series:
"SeasonallyAdjusted": after the decomposition the effect is allocated to
the seasonally adjusted series:
More information on outliers and other auxiliary variables in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/
set_tradingdays, intervention_variable
# creating one or several external regressors (TS objects), # which will be gathered in one or several groups iv1 <- intervention_variable(12, c(2000, 1), 60, starts = "2001-01-01", ends = "2001-12-01" ) iv2 <- intervention_variable(12, c(2000, 1), 60, starts = "2001-01-01", ends = "2001-12-01", delta = 1 ) # configuration 1: regressors in the same default group (named "r") variables <- list("iv1" = iv1, "iv2" = iv2) # to use those regressors, input : name=r.iv1 and r.iv2 in add_usrdefvar function # configuration 2: group names are user-defined # here: regressors as a list of two groups (lists) reg1 and reg2 vars <- list(reg1 = list(iv1 = iv1), reg2 = list(iv2 = iv2)) # to use those regressors, input : name=reg1.iv1 and name=reg2.iv2 in add_usrdefvar function # creating the modelling context my_context <- modelling_context(variables = vars) # customize a default specification # init_spec <- rjd3x13::x13_spec("RSA5c") # regressors have to be added one by one # new_spec<- add_usrdefvar(init_spec,name = "reg1.iv1", regeffect="Trend") # new spec<- add_usrdefvar(new_spec,name = "reg2.iv2", regeffect="Trend", coef=0.7) # modelling context is needed for the estimation phase # sa_x13<- rjd3x13::x13(ABS$X0.2.09.10.M, new_spec, context = my_context)# creating one or several external regressors (TS objects), # which will be gathered in one or several groups iv1 <- intervention_variable(12, c(2000, 1), 60, starts = "2001-01-01", ends = "2001-12-01" ) iv2 <- intervention_variable(12, c(2000, 1), 60, starts = "2001-01-01", ends = "2001-12-01", delta = 1 ) # configuration 1: regressors in the same default group (named "r") variables <- list("iv1" = iv1, "iv2" = iv2) # to use those regressors, input : name=r.iv1 and r.iv2 in add_usrdefvar function # configuration 2: group names are user-defined # here: regressors as a list of two groups (lists) reg1 and reg2 vars <- list(reg1 = list(iv1 = iv1), reg2 = list(iv2 = iv2)) # to use those regressors, input : name=reg1.iv1 and name=reg2.iv2 in add_usrdefvar function # creating the modelling context my_context <- modelling_context(variables = vars) # customize a default specification # init_spec <- rjd3x13::x13_spec("RSA5c") # regressors have to be added one by one # new_spec<- add_usrdefvar(init_spec,name = "reg1.iv1", regeffect="Trend") # new spec<- add_usrdefvar(new_spec,name = "reg2.iv2", regeffect="Trend", coef=0.7) # modelling context is needed for the estimation phase # sa_x13<- rjd3x13::x13(ABS$X0.2.09.10.M, new_spec, context = my_context)
Makes a frequency change of this series.
aggregate( s, nfreq = 1, conversion = c("Sum", "Average", "First", "Last", "Min", "Max"), complete = TRUE )aggregate( s, nfreq = 1, conversion = c("Sum", "Average", "First", "Last", "Min", "Max"), complete = TRUE )
s |
the input time series. |
nfreq |
the new frequency. Must be la divisor of the frequency of |
conversion |
Aggregation mode: sum ( |
complete |
Boolean indicating if the observation for a given period in the new series is set missing if some data in the original series are missing. |
A new time series of frequency nfreq.
s <- ABS$X0.2.09.10.M # Annual sum aggregate(s, nfreq = 1, conversion = "Sum") # first and last years removed aggregate(s, nfreq = 1, conversion = "Sum", complete = FALSE) # Quarterly mean aggregate(s, nfreq = 4, conversion = "Average")s <- ABS$X0.2.09.10.M # Annual sum aggregate(s, nfreq = 1, conversion = "Sum") # first and last years removed aggregate(s, nfreq = 1, conversion = "Sum", complete = FALSE) # Quarterly mean aggregate(s, nfreq = 4, conversion = "Average")
Remove an arima model from an existing one. More exactly, m_diff = m_left - m_right iff m_left = m_right + m_diff.
arima_difference(left, right, simplify = TRUE)arima_difference(left, right, simplify = TRUE)
left |
Left operand (JD3_ARIMA object) |
right |
Right operand (JD3_ARIMA object) |
simplify |
Simplify the results if possible (common roots in the auto-regressive and in the moving average polynomials, including unit roots) |
a "JD3_ARIMA" model.
mod1 <- arima_model(delta = c(1, -2, 1)) mod2 <- arima_model(variance = .01) diff <- arima_difference(mod1, mod2) sum <- arima_sum(diff, mod2) # sum should be equal to mod1mod1 <- arima_model(delta = c(1, -2, 1)) mod2 <- arima_model(variance = .01) diff <- arima_difference(mod1, mod2) sum <- arima_sum(diff, mod2) # sum should be equal to mod1
ARIMA Model
arima_model(name = "arima", ar = 1, delta = 1, ma = 1, variance = 1)arima_model(name = "arima", ar = 1, delta = 1, ma = 1, variance = 1)
name |
Name of the model. |
ar |
coefficients of the regular auto-regressive polynomial (1 + ar(1)B + ar(2)B + ...). True signs. |
delta |
non stationary auto-regressive polynomial. |
ma |
coefficients of the regular moving average polynomial (1 + ma(1)B + ma(2)B + ...). True signs. |
variance |
variance. |
a "JD3_ARIMA" model.
model <- arima_model("trend", ar = c(1, -.8), delta = c(1, -1), ma = c(1, -.5), var = 100)model <- arima_model("trend", ar = c(1, -.8), delta = c(1, -1), ma = c(1, -.5), var = 100)
Properties of an ARIMA model; the (pseudo-)spectrum and the auto-covariances of the model are returned
arima_properties(model, nspectrum = 601, nac = 36)arima_properties(model, nspectrum = 601, nac = 36)
model |
a |
nspectrum |
number of points to calculate the spectrum; th points are uniformly distributed in [0, pi] |
nac |
maximum lag at which to calculate the auto-covariances; if the model is non-stationary, the auto-covariances are computed on its stationary transformation. |
A list with tha auto-covariances and with the (pseudo-)spectrum
mod1 <- arima_model(ar = c(0.1, 0.2), delta = c(1, -1), ma = 0) arima_properties(mod1)mod1 <- arima_model(ar = c(0.1, 0.2), delta = c(1, -1), ma = 0) arima_properties(mod1)
Sum ARIMA Models
arima_sum(...)arima_sum(...)
... |
list of ARIMA models (created with |
Adds several Arima models, considering that their innovations are independent. The sum of two Arima models is computed as follows: the auto-regressive parts (stationary and non stationary of the aggregated model are the smaller common multiple of the corresponding polynomials of the components. The sum of the acf of the modified moving average polynomials is then computed and factorized, to get the moving average polynomial and innovation variance of the sum.
a "JD3_ARIMA" model.
mod1 <- arima_model(ar = c(0.1, 0.2), delta = 0, ma = 0) mod2 <- arima_model(ar = 0, delta = 0, ma = c(0.4)) arima_sum(mod1, mod2)mod1 <- arima_model(ar = c(0.1, 0.2), delta = 0, ma = 0) mod2 <- arima_model(ar = 0, delta = 0, ma = c(0.4)) arima_sum(mod1, mod2)
Autocorrelation Functions
autocorrelations(data, mean = TRUE, n = 15) autocorrelations_partial(data, mean = TRUE, n = 15) autocorrelations_inverse(data, nar = 30, n = 15)autocorrelations(data, mean = TRUE, n = 15) autocorrelations_partial(data, mean = TRUE, n = 15) autocorrelations_inverse(data, nar = 30, n = 15)
data |
data being tested. |
mean |
Mean correction. If |
n |
maximum lag at which to calculate the stats. |
nar |
number of AR lags used to compute inverse autocorrelations. |
x <- ABS$X0.2.09.10.M autocorrelations(x) autocorrelations_partial(x) autocorrelations_inverse(x)x <- ABS$X0.2.09.10.M autocorrelations(x) autocorrelations_partial(x) autocorrelations_inverse(x)
Allows to generate trading day regressors (as many as defined groups), taking into account
7 or less different types of days, from Monday to Sunday, and specific holidays,which are to
defined beforehand in a calendar using the functions national_calendar,weighted_calendar or
Chained_calendar.
calendar_td( calendar, frequency, start, length, s, groups = c(1, 2, 3, 4, 5, 6, 0), holiday = 7, contrasts = TRUE )calendar_td( calendar, frequency, start, length, s, groups = c(1, 2, 3, 4, 5, 6, 0), holiday = 7, contrasts = TRUE )
calendar |
The calendar containing the required holidays |
frequency |
Frequency of the series, number of periods per year (12,4,3,2..) |
start, length
|
First date (array with the first year and the first period)
(for instance |
s |
time series used to get the dates for the trading days variables. If supplied the
parameters |
groups |
Groups of days. The length of the array must be 7. It indicates to what group each week day belongs. The first item corresponds to Mondays and the last one to Sundays. The group used for contrasts (usually Sundays) is identified by 0. The other groups are identified by 1, 2,... n (<= 6). For instance, usual trading days are defined by c(1,2,3,4,5,6,0), week days by c(1,1,1,1,1,0,0), week days, Saturdays, Sundays by c(1,1,1,1,1,2,0) etc. |
holiday |
Day to aggregate holidays with. (holidays are considered as that day). 1 for Monday... 7 for Sunday. Doesn't necessary belong to the 0-group. |
contrasts |
If true, the variables are defined by contrasts with the 0-group. Otherwise, raw number of days is provided. |
Aggregated values for monthly or quarterly are the numbers of days belonging to a given group, holidays
are all summed together in of those groups.
Contrasts are the differences between the number of days in a given group (1 to 6) and the number of days in
the reference group (0).
Regressors are corrected for long-term mean if contrasts = TRUE.
Time series (object of class c("ts","mts","matrix")) corresponding to each group, starting with the 0-group (contrasts = FALSE)
or the 1-group (contrasts = TRUE).
More information on calendar correction in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/
BE <- national_calendar(list( fixed_day(7, 21), special_day("NEWYEAR"), special_day("CHRISTMAS"), special_day("MAYDAY"), special_day("EASTERMONDAY"), special_day("ASCENSION"), special_day("WHITMONDAY"), special_day("ASSUMPTION"), special_day("ALLSAINTSDAY"), special_day("ARMISTICE") )) calendar_td(BE, 12, c(1980, 1), 240, holiday = 7, groups = c(1, 1, 1, 2, 2, 3, 0), contrasts = FALSE )BE <- national_calendar(list( fixed_day(7, 21), special_day("NEWYEAR"), special_day("CHRISTMAS"), special_day("MAYDAY"), special_day("EASTERMONDAY"), special_day("ASCENSION"), special_day("WHITMONDAY"), special_day("ASSUMPTION"), special_day("ALLSAINTSDAY"), special_day("ARMISTICE") )) calendar_td(BE, 12, c(1980, 1), 240, holiday = 7, groups = c(1, 1, 1, 2, 2, 3, 0), contrasts = FALSE )
Allows to combine two calendars, one before and one after a given date.
chained_calendar(calendar1, calendar2, break_date)chained_calendar(calendar1, calendar2, break_date)
calendar1, calendar2
|
calendars to chain. |
break_date |
the break date in the format |
A chained calendar is an useful option when major changes in the composition of the holidays take place. In such a case two calendars describing the situation before and after the change of regime can be defined and bound together, one before the break and one after the break.
returns an object of class c("JD3_CHAINEDCALENDAR","JD3_CALENDARDEFINITION")
More information on calendar correction in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/a-calendar-correction
national_calendar, weighted_calendar
Belgium <- national_calendar(list(special_day("NEWYEAR"), fixed_day(7, 21))) France <- national_calendar(list(special_day("NEWYEAR"), fixed_day(7, 14))) chained_cal <- chained_calendar(France, Belgium, "2000-01-01")Belgium <- national_calendar(list(special_day("NEWYEAR"), fixed_day(7, 21))) France <- national_calendar(list(special_day("NEWYEAR"), fixed_day(7, 14))) chained_cal <- chained_calendar(France, Belgium, "2000-01-01")
Removal of missing values at the beginning/end
clean_extremities(s)clean_extremities(s)
s |
Original series |
Cleaned series
y <- window(ABS$X0.2.09.10.M, start = 1982, end = 2018, extend = TRUE) y clean_extremities(y)y <- window(ABS$X0.2.09.10.M, start = 1982, end = 2018, extend = TRUE) y clean_extremities(y)
Compare the annual totals of two series (usually the raw series and the seasonally adjusted series)
compare_annual_totals(raw, sa)compare_annual_totals(raw, sa)
raw |
Raw series |
sa |
Seasonally adjusted series |
The largest annual difference (in percentage of the average level of the seasonally adjusted series)
ts of JDemetra+Promote a R time series to a "full" ts of JDemetra+
data_to_ts(s, name)data_to_ts(s, name)
s |
R time series |
name |
name of the series |
s <- ABS$X0.2.09.10.M t <- data_to_ts(s, "test")s <- ABS$X0.2.09.10.M t <- data_to_ts(s, "test")
Provides a list of dates corresponding to each period of the given time series
daysOf(ts, pos = 1)daysOf(ts, pos = 1)
ts |
A time series |
pos |
The position of the first considered period. |
A list of the starting dates of each period
daysOf(retail$BookStores)daysOf(retail$BookStores)
Density, (cumulative) distribution function and random generation for chi-squared distribution.
density_chi2(df, x) cdf_chi2(df, x) random_chi2(df, n)density_chi2(df, x) cdf_chi2(df, x) random_chi2(df, n)
df |
degrees of freedom. |
x |
vector of quantiles. |
n |
number of observations. |
Density, (cumulative) distribution function and random generation for Gamma distribution.
density_gamma(shape, scale, x) cdf_gamma(shape, scale, x) random_gamma(shape, scale, n)density_gamma(shape, scale, x) cdf_gamma(shape, scale, x) random_gamma(shape, scale, n)
shape, scale
|
shape and scale parameters. |
x |
vector of quantiles. |
n |
number of observations. |
Density, (cumulative) distribution function and random generation for inverse-gamma distribution.
density_inverse_gamma(shape, scale, x) cdf_inverse_gamma(shape, scale, x) random_inverse_gamma(shape, scale, n)density_inverse_gamma(shape, scale, x) cdf_inverse_gamma(shape, scale, x) random_inverse_gamma(shape, scale, n)
shape, scale
|
shape and scale parameters. |
x |
vector of quantiles. |
n |
number of observations. |
Density, (cumulative) distribution function and random generation for inverse-gaussian distribution.
density_inverse_gaussian(shape, scale, x) cdf_inverse_gaussian(shape, scale, x) random_inverse_gaussian(shape, scale, n)density_inverse_gaussian(shape, scale, x) cdf_inverse_gaussian(shape, scale, x) random_inverse_gaussian(shape, scale, n)
shape, scale
|
shape and scale parameters. |
x |
vector of quantiles. |
n |
number of observations. |
Density, (cumulative) distribution function and random generation for Student distribution.
density_t(df, x) cdf_t(df, x) random_t(df, n)density_t(df, x) cdf_t(df, x) random_t(df, n)
df |
degrees of freedom. |
x |
vector of quantiles. |
n |
number of observations. |
# T with 2 degrees of freedom. z <- density_t(2, .01 * seq(-100, 100, 1)) # T with 2 degrees of freedom. 100 random z <- random_t(2, 100)# T with 2 degrees of freedom. z <- density_t(2, .01 * seq(-100, 100, 1)) # T with 2 degrees of freedom. 100 random z <- random_t(2, 100)
Use sa_decomposition() instead of sa.decomposition().
sa.decomposition(x, ...)sa.decomposition(x, ...)
x |
the object to print. |
... |
further arguments. |
Generic Diagnostics Function
diagnostics(x, ...) ## S3 method for class 'JD3' diagnostics(x, ...)diagnostics(x, ...) ## S3 method for class 'JD3' diagnostics(x, ...)
x |
the object to extract diagnostics. |
... |
further arguments. |
Extract dictionary of a "JD3_ProcResults" object (dictionary()) and extract a specific value (result()) or a list of values (user_defined()).
dictionary(object) result(object, id) user_defined(object, userdefined = NULL)dictionary(object) result(object, id) user_defined(object, userdefined = NULL)
object |
the java object. |
id |
the name of the object to extract. |
userdefined |
vector containing the names of the object to extract. |
Differencing of a series
differences(data, lags = 1, mean = TRUE)differences(data, lags = 1, mean = TRUE)
data |
The series to be differenced. |
lags |
Lags of the differencing. |
mean |
Apply a mean correction at the end of the differencing process. |
The differenced series.
differences(retail$BookStores, c(1, 1, 12), FALSE)differences(retail$BookStores, c(1, 1, 12), FALSE)
The series is differenced till its variance is decreasing.
differencing_fast(data, period, mad = TRUE, centile = 90, k = 1.2)differencing_fast(data, period, mad = TRUE, centile = 90, k = 1.2)
data |
Series being differenced. |
period |
Period considered in the automatic differencing. |
mad |
Use of MAD in the computation of the variance (true by default). |
centile |
Percentage of the data used for computing the variance (90 by default). |
k |
tolerance in the decrease of the variance. The algorithm stops if the new variance is higher than k*the old variance. k should be equal or slightly higher than 1 (1.2 by default) |
Stationary transformation
ddata: data after differencing
mean: mean correction
differences:
lag:
order: order of the differencing
differencing_fast(log(ABS$X0.2.09.10.M), 12)differencing_fast(log(ABS$X0.2.09.10.M), 12)
Automatic processing (identification of the order of the differencing) based on auto-correlations and on mean correction. The series should not be seasonal. Source: Tramo
do_stationary(data, period)do_stationary(data, period)
data |
Series being differenced. |
period |
Period of the series. |
Stationary transformation
ddata: data after differencing
mean: mean correction
differences:
lag:
order: order of the differencing
do_stationary(log(ABS$X0.2.09.10.M), 12)do_stationary(log(ABS$X0.2.09.10.M), 12)
Allows to display the date of Easter Sunday for each year, in the defined period. Dates are displayed in "YYYY-MM-DD" format and as a number of days since January 1st 1970.
easter_dates(year0, year1, julian = FALSE)easter_dates(year0, year1, julian = FALSE)
year0, year1
|
starting year and ending year |
julian |
Boolean indicating if Julian calendar must be used. |
a named numeric vector. Names are the dates in format "YYYY-MM-DD", values are number of days since January 1st 1970.
More information on calendar correction in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/a-calendar-correction
# Dates from 2018(included) to 2023 (included) easter_dates(2018, 2023)# Dates from 2018(included) to 2023 (included) easter_dates(2018, 2023)
Allows to define a holiday which date is related to Easter Sunday.
easter_day(offset, julian = FALSE, weight = 1, validity = NULL)easter_day(offset, julian = FALSE, weight = 1, validity = NULL)
offset |
The position of the holiday in relation to Easter Sunday, measured in days (can be positive or negative). |
julian |
Boolean indicating if Julian calendar must be used. |
weight |
weight associated to the holiday. |
validity |
validity period: either |
More information on calendar correction in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/a-calendar-correction
national_calendar, fixed_day,special_day,fixed_week_day
easter_day(1) # Easter Monday easter_day(-2) # Easter Good Friday # Corpus Christi 60 days after Easter # Sunday in Julian calendar with weight 0.5, from January 2000 to December 2020 easter_day( offset = 60, julian = TRUE, weight = 0.5, validity = list(start = "2000-01-01", end = "2020-12-01") )easter_day(1) # Easter Monday easter_day(-2) # Easter Good Friday # Corpus Christi 60 days after Easter # Sunday in Julian calendar with weight 0.5, from January 2000 to December 2020 easter_day( offset = 60, julian = TRUE, weight = 0.5, validity = list(start = "2000-01-01", end = "2020-12-01") )
Allows to generate a regressor taking into account the (Julian) Easter effect in monthly or quarterly time series.
easter_variable( frequency, start, length, s, duration = 6, endpos = -1, correction = c("Simple", "PreComputed", "Theoretical", "None") ) julianeaster_variable(frequency, start, length, s, duration = 6)easter_variable( frequency, start, length, s, duration = 6, endpos = -1, correction = c("Simple", "PreComputed", "Theoretical", "None") ) julianeaster_variable(frequency, start, length, s, duration = 6)
frequency |
Frequency of the series, number of periods per year (12,4,3,2..) |
start, length
|
First date (array with the first year and the first period)
(for instance |
s |
time series used to get the dates for the trading days variables. If supplied the
parameters |
duration |
Duration (length in days) of the Easter effect. (value between 1 and 20, default =6) |
endpos |
Position of the end of the Easter effect, relatively to Easter: -1(default): before Easter Sunday, 0: on Easter Sunday, 1: on Easter Monday) |
correction |
mean correction option. Simple"(default), "PreComputed", "Theoretical" or "None". |
A time series (object of class "ts")
More information on calendar correction in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/a-calendar-correction
# Monthly regressor, five-year long, duration 8 days, effect finishing on Easter Monday ee <- easter_variable(12, c(2020, 1), length = 5 * 12, duration = 8, endpos = 1)# Monthly regressor, five-year long, duration 8 days, effect finishing on Easter Monday ee <- easter_variable(12, c(2020, 1), length = 5 * 12, duration = 8, endpos = 1)
Belgian exports to European countries
ExportsExports
An object of class list of length 34.
NBB
creates a holiday falling on a fixed day each year, with an optional weight and period of validity, like Christmas which is always celebrated on December 25th.
fixed_day(month, day, weight = 1, validity = NULL)fixed_day(month, day, weight = 1, validity = NULL)
month, day
|
the month and the day of the fixed day to add. |
weight |
weight associated to the holiday. |
validity |
validity period: either |
returns an object of class c("JD3_FIXEDDAY","JD3_HOLIDAY")
More information on calendar correction in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/a-calendar-correction
national_calendar, special_day,easter_day
day <- fixed_day(7, 21, .9) day # July 21st, with weight=0.9, on the whole sample day <- fixed_day(12, 25, .5, validity = list(start = "2010-01-01")) day # December 25th, with weight=0.5, from January 2010 day <- fixed_day(12, 25, .5, validity = list(start = "1968-02-01", end = "2010-01-01")) day # December 25th, with weight=0.9, from February 1968 until January 2010day <- fixed_day(7, 21, .9) day # July 21st, with weight=0.9, on the whole sample day <- fixed_day(12, 25, .5, validity = list(start = "2010-01-01")) day # December 25th, with weight=0.5, from January 2010 day <- fixed_day(12, 25, .5, validity = list(start = "1968-02-01", end = "2010-01-01")) day # December 25th, with weight=0.9, from February 1968 until January 2010
Allows to define a holiday falling on a fixed week day each year, like Labour Day in the USA which is always celebrated on the first Monday of September.
fixed_week_day(month, week, dayofweek, weight = 1, validity = NULL)fixed_week_day(month, week, dayofweek, weight = 1, validity = NULL)
month |
month of the holiday: from |
week |
position of the specified week day in the month: from |
dayofweek |
day of the week: from |
weight |
weight associated to the holiday. |
validity |
validity period: either |
returns an object of class c("JD3_FIXEDWEEKDAY","JD3_HOLIDAY")
More information on calendar correction in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/a-calendar-correction
national_calendar, fixed_day,special_day,easter_day
day <- fixed_week_day(9, 1, 1) # first Monday(1) of September. dayday <- fixed_week_day(9, 1, 1) # first Monday(1) of September. day
Allows to generate daily regressors (dummy variables) corresponding to each holiday of a pre-defined calendar.
holidays( calendar, start, length, nonworking = c(6, 7), type = c("Skip", "All", "NextWorkingDay", "PreviousWorkingDay"), single = FALSE )holidays( calendar, start, length, nonworking = c(6, 7), type = c("Skip", "All", "NextWorkingDay", "PreviousWorkingDay"), single = FALSE )
calendar |
The calendar in which the holidays are defined. |
start |
Starting date for the regressors, format |
length |
Length of the regressors in days. |
nonworking |
Indexes of non working days (Monday=1, Sunday=7). |
type |
Adjustment type when a holiday falls on a week-end:
|
single |
Boolean indication if a single variable ( |
The pre-defined in a calendar has to be created with the functions national_calendar or weighted_calendar or
weighted_calendar. A many regressors as defined holidays are generated, when the holiday occurs
the value is 1 and 0 otherwise.
This kind of non-aggregated regressors are used for calendar correction in daily data.
A matrix (class "matrix") where each column is associated to a holiday (in the order of creation of the holiday) and each row to a date.
More information on calendar correction in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/a-calendar-correction
BE <- national_calendar(list( fixed_day(7, 21), special_day("NEWYEAR"), special_day("CHRISTMAS"), special_day("MAYDAY"), special_day("EASTERMONDAY"), special_day("ASCENSION"), special_day("WHITMONDAY"), special_day("ASSUMPTION"), special_day("ALLSAINTSDAY"), special_day("ARMISTICE") )) q <- holidays(BE, "2021-01-01", 366 * 10, type = "All") plot(apply(q, 1, max))BE <- national_calendar(list( fixed_day(7, 21), special_day("NEWYEAR"), special_day("CHRISTMAS"), special_day("MAYDAY"), special_day("EASTERMONDAY"), special_day("ASCENSION"), special_day("WHITMONDAY"), special_day("ASSUMPTION"), special_day("ALLSAINTSDAY"), special_day("ARMISTICE") )) q <- holidays(BE, "2021-01-01", 366 * 10, type = "All") plot(apply(q, 1, max))
Belgian imports from European countries
ImportsImports
An object of class list of length 34.
NBB
Function allowing to create external regressors as sequences of zeros and ones. The generated variables
will have to be added with add_usrdefvar function will require a modelling context definition
with modelling_context to be used in an estimation process.
intervention_variable( frequency, start, length, s, starts, ends, delta = 0, seasonaldelta = 0 )intervention_variable( frequency, start, length, s, starts, ends, delta = 0, seasonaldelta = 0 )
frequency |
Frequency of the series, number of periods per year (12,4,3,2..) |
start, length
|
First date (array with the first year and the first period)
(for instance |
s |
time series used to get the dates for the trading days variables. If supplied the
parameters |
starts, ends
|
characters specifying sequences of starts/ends dates for the intervention variable. Can be characters or integers. |
delta |
regular differencing order. |
seasonaldelta |
seasonal differencing order. |
Intervention variables are combinations of any possible sequence of ones and zeros
(the sequence of ones being defined by the parameters starts and ends)
and of and where is the
backwards operator, is the frequency of the time series,
and are the parameters delta and seasonaldelta.
For example, with delta = 0 and seasonaldelta = 0 we get temporary level shifts defined
by the parameters starts and ends. With delta = 1 and seasonaldelta = 0 we get
the cumulative sum of temporary level shifts, once differenced the regressor will become a classical level shift.
More information on auxiliary variables in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/
modelling_context, add_usrdefvar
iv1 <- intervention_variable(12, c(2000, 1), 60, starts = "2001-01-01", ends = "2001-12-01" ) plot(iv1) iv2 <- intervention_variable(12, c(2000, 1), 60, starts = "2001-01-01", ends = "2001-12-01", delta = 1 ) plot(iv2) # using one variable in a a seasonal adjustment process # regressors as a list of two groups reg1 and reg2 vars <- list(reg1 = list(x = iv1), reg2 = list(x = iv2)) # creating the modelling context my_context <- modelling_context(variables = vars) # customize a default specification # init_spec <- rjd3x13::x13_spec("RSA5c") # new_spec<- add_usrdefvar(init_spec,id = "reg1.iv1", regeffect="Trend") # modelling context is needed for the estimation phase # sa_x13<- rjd3x13::x13(ABS$X0.2.09.10.M, new_spec, context = my_context)iv1 <- intervention_variable(12, c(2000, 1), 60, starts = "2001-01-01", ends = "2001-12-01" ) plot(iv1) iv2 <- intervention_variable(12, c(2000, 1), 60, starts = "2001-01-01", ends = "2001-12-01", delta = 1 ) plot(iv2) # using one variable in a a seasonal adjustment process # regressors as a list of two groups reg1 and reg2 vars <- list(reg1 = list(x = iv1), reg2 = list(x = iv2)) # creating the modelling context my_context <- modelling_context(variables = vars) # customize a default specification # init_spec <- rjd3x13::x13_spec("RSA5c") # new_spec<- add_usrdefvar(init_spec,id = "reg1.iv1", regeffect="Trend") # modelling context is needed for the estimation phase # sa_x13<- rjd3x13::x13(ABS$X0.2.09.10.M, new_spec, context = my_context)
JD3 print functions
## S3 method for class 'JD3_ARIMA' print(x, ...) ## S3 method for class 'JD3_UCARIMA' print(x, ...) ## S3 method for class 'JD3_SARIMA' print(x, ...) ## S3 method for class 'JD3_SARIMA_ESTIMATION' print(x, digits = max(3L, getOption("digits") - 3L), ...) ## S3 method for class 'JD3_SPAN' print(x, ...) ## S3 method for class 'JD3_LIKELIHOOD' print(x, ...) ## S3 method for class 'JD3_REGARIMA_RSLTS' print( x, digits = max(3L, getOption("digits") - 3L), summary_info = getOption("summary_info"), ... )## S3 method for class 'JD3_ARIMA' print(x, ...) ## S3 method for class 'JD3_UCARIMA' print(x, ...) ## S3 method for class 'JD3_SARIMA' print(x, ...) ## S3 method for class 'JD3_SARIMA_ESTIMATION' print(x, digits = max(3L, getOption("digits") - 3L), ...) ## S3 method for class 'JD3_SPAN' print(x, ...) ## S3 method for class 'JD3_LIKELIHOOD' print(x, ...) ## S3 method for class 'JD3_REGARIMA_RSLTS' print( x, digits = max(3L, getOption("digits") - 3L), summary_info = getOption("summary_info"), ... )
x |
the object to print. |
... |
further unused parameters. |
digits |
minimum number of significant digits to be used for most numbers. |
summary_info |
boolean indicating if a message suggesting the use of the
summary function for more details should be printed. By default used the
option |
Compute Ljung-Box test to check the independence of a data.
ljungbox(data, k = 1, lag = 1, nhp = 0, sign = 0, mean = TRUE)ljungbox(data, k = 1, lag = 1, nhp = 0, sign = 0, mean = TRUE)
data |
data being tested. |
k |
number of auto-correlations used in the test |
lag |
number of lags used between two auto-correlations. |
nhp |
number of hyper parameters (to correct the degree of freedom) |
sign |
if |
mean |
Mean correction. If |
A c("JD3_TEST", "JD3") object (see statisticaltest() for details).
ljungbox(random_t(2, 100), lag = 24, k = 1) ljungbox(ABS$X0.2.09.10.M, lag = 24, k = 1)ljungbox(random_t(2, 100), lag = 24, k = 1) ljungbox(ABS$X0.2.09.10.M, lag = 24, k = 1)
Given a pre-defined calendar and set of groups, the function displays the long-term means which would be used to seasonally adjust the corresponding regressors, as the final value using contrasts is "number of days in the group - long term mean".
long_term_mean( calendar, frequency, groups = c(1, 2, 3, 4, 5, 6, 0), holiday = 7 )long_term_mean( calendar, frequency, groups = c(1, 2, 3, 4, 5, 6, 0), holiday = 7 )
calendar |
The calendar containing the required holidays |
frequency |
Frequency of the series, number of periods per year (12,4,3,2..) |
groups |
Groups of days. The length of the array must be 7. It indicates to what group each week day belongs. The first item corresponds to Mondays and the last one to Sundays. The group used for contrasts (usually Sundays) is identified by 0. The other groups are identified by 1, 2,... n (<= 6). For instance, usual trading days are defined by c(1,2,3,4,5,6,0), week days by c(1,1,1,1,1,0,0), week days, Saturdays, Sundays by c(1,1,1,1,1,2,0) etc. |
holiday |
Day to aggregate holidays with. (holidays are considered as that day). 1 for Monday... 7 for Sunday. Doesn't necessary belong to the 0-group. |
A long-term mean is a probability based computation of the average value for every period in every group. (see references). For monthly regressors there are 12 types of periods (January to December).
returns an object of class c("matrix","array") with the long term means corresponding
to each group/period, starting with the 0-group.
BE <- national_calendar(list( fixed_day(7, 21), special_day("NEWYEAR"), special_day("CHRISTMAS"), special_day("MAYDAY"), special_day("EASTERMONDAY"), special_day("ASCENSION"), special_day("WHITMONDAY"), special_day("ASSUMPTION"), special_day("ALLSAINTSDAY"), special_day("ARMISTICE") )) lt <- long_term_mean(BE, 12, groups = c(1, 1, 1, 1, 1, 0, 0), holiday = 7 )BE <- national_calendar(list( fixed_day(7, 21), special_day("NEWYEAR"), special_day("CHRISTMAS"), special_day("MAYDAY"), special_day("EASTERMONDAY"), special_day("ASCENSION"), special_day("WHITMONDAY"), special_day("ASSUMPTION"), special_day("ALLSAINTSDAY"), special_day("ARMISTICE") )) lt <- long_term_mean(BE, 12, groups = c(1, 1, 1, 1, 1, 0, 0), holiday = 7 )
Allows to generate a regressor correcting for the leap year or length-of-period effect.
lp_variable( frequency, start, length, s, type = c("LeapYear", "LengthOfPeriod") )lp_variable( frequency, start, length, s, type = c("LeapYear", "LengthOfPeriod") )
frequency |
Frequency of the series, number of periods per year (12,4,3,2..) |
start, length
|
First date (array with the first year and the first period)
(for instance |
s |
time series used to get the dates for the trading days variables. If supplied the
parameters |
type |
the modelling of the leap year effect: as a contrast variable ( |
Time series (object of class "ts")
More information on calendar correction in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/a-calendar-correction
# Leap years occur in year 2000, 2004, 2008 and 2012 lp_variable(4, start = c(2000, 1), length = 4 * 13) lper <- lp_variable(12, c(2000, 1), length = 10 * 12, type = "LengthOfPeriod")# Leap years occur in year 2000, 2004, 2008 and 2012 lp_variable(4, start = c(2000, 1), length = 4 * 13) lper <- lp_variable(12, c(2000, 1), length = 10 * 12, type = "LengthOfPeriod")
Compute a robust median absolute deviation (MAD)
mad(data, centile = 50, medianCorrected = TRUE)mad(data, centile = 50, medianCorrected = TRUE)
data |
The data for which we compute the robust deviation |
centile |
The centile used to exclude extreme values (only the "centile" part of the data are is to compute the mad) |
medianCorrected |
TRUE if the series is corrected for its median, FALSE if the median is supposed to be 0 |
The median absolute deviation
y <- rnorm(1000) m <- rjd3toolkit::mad(y, centile = 70)y <- rnorm(1000) m <- rjd3toolkit::mad(y, centile = 70)
Function allowing to include calendars and external regressors in a format that makes them usable
in an estimation processes (seasonal adjustment or pre-processing). The regressors can be created with functions available in the package
or come from any other source, provided they are ts class objects.
modelling_context(calendars = NULL, variables = NULL)modelling_context(calendars = NULL, variables = NULL)
calendars |
list of calendars. |
variables |
list of variables. |
list of calendars and variables
More information on auxiliary variables in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/
add_usrdefvar, intervention_variable
# creating one or several external regressors (TS objects), which will # be gathered in one or several groups iv1 <- intervention_variable(12, c(2000, 1), 60, starts = "2001-01-01", ends = "2001-12-01" ) iv2 <- intervention_variable(12, c(2000, 1), 60, starts = "2001-01-01", ends = "2001-12-01", delta = 1 ) # regressors as a list of two groups reg1 and reg2 vars <- list(reg1 = list(x = iv1), reg2 = list(x = iv2)) # creating the modelling context my_context <- modelling_context(variables = vars) # customize a default specification # init_spec <- rjd3x13::x13_spec("RSA5c") # new_spec<- add_usrdefvar(init_spec,name = "reg1.iv1", regeffect="Trend") # modelling context is needed for the estimation phase # sa_x13<- rjd3x13::x13(ABS$X0.2.09.10.M, new_spec, context = my_context)# creating one or several external regressors (TS objects), which will # be gathered in one or several groups iv1 <- intervention_variable(12, c(2000, 1), 60, starts = "2001-01-01", ends = "2001-12-01" ) iv2 <- intervention_variable(12, c(2000, 1), 60, starts = "2001-01-01", ends = "2001-12-01", delta = 1 ) # regressors as a list of two groups reg1 and reg2 vars <- list(reg1 = list(x = iv1), reg2 = list(x = iv2)) # creating the modelling context my_context <- modelling_context(variables = vars) # customize a default specification # init_spec <- rjd3x13::x13_spec("RSA5c") # new_spec<- add_usrdefvar(init_spec,name = "reg1.iv1", regeffect="Trend") # modelling context is needed for the estimation phase # sa_x13<- rjd3x13::x13(ABS$X0.2.09.10.M, new_spec, context = my_context)
Will create a calendar as a list of days corresponding to the required holidays.
The holidays have to be generated by one of these functions: fixed_day(),
fixed_week_day(), easter_day(), special_day() or single_day().
national_calendar(days, mean_correction = TRUE)national_calendar(days, mean_correction = TRUE)
days |
list of holidays to be taken into account in the calendar |
mean_correction |
TRUE if the variables generated by this calendar will contain long term mean corrections (default). FALSE otherwise. |
returns an object of class c("JD3_CALENDAR","JD3_CALENDARDEFINITION")
More information on calendar correction in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/
chained_calendar, weighted_calendar
# Fictional calendar using all possibilities to set the required holidays MyCalendar <- national_calendar(list( fixed_day(7, 21), special_day("NEWYEAR"), special_day("CHRISTMAS"), fixed_week_day(7, 2, 3), # second Wednesday of July special_day("MAYDAY"), easter_day(1), # Easter Monday easter_day(-2), # Good Friday single_day("2001-09-11"), # appearing once special_day("ASCENSION"), easter_day( offset = 60, julian = FALSE, weight = 0.5, validity = list(start = "2000-01-01", end = "2020-12-01") ), # Corpus Christi special_day("WHITMONDAY"), special_day("ASSUMPTION"), special_day("ALLSAINTSDAY"), special_day("ARMISTICE") ))# Fictional calendar using all possibilities to set the required holidays MyCalendar <- national_calendar(list( fixed_day(7, 21), special_day("NEWYEAR"), special_day("CHRISTMAS"), fixed_week_day(7, 2, 3), # second Wednesday of July special_day("MAYDAY"), easter_day(1), # Easter Monday easter_day(-2), # Good Friday single_day("2001-09-11"), # appearing once special_day("ASCENSION"), easter_day( offset = 60, julian = FALSE, weight = 0.5, validity = list(start = "2000-01-01", end = "2020-12-01") ), # Corpus Christi special_day("WHITMONDAY"), special_day("ASSUMPTION"), special_day("ALLSAINTSDAY"), special_day("ARMISTICE") ))
Set of functions to test the normality of a time series.
bowmanshenton(data) doornikhansen(data) jarquebera(data, k = 0, sample = TRUE) skewness(data) kurtosis(data)bowmanshenton(data) doornikhansen(data) jarquebera(data, k = 0, sample = TRUE) skewness(data) kurtosis(data)
data |
data being tested. |
k |
number of degrees of freedom to be subtracted if the input time series is a series of residuals. |
sample |
boolean indicating if unbiased empirical moments should be computed. |
A c("JD3_TEST", "JD3") object (see statisticaltest for details).
bowmanshenton(): Bowman-Shenton test
doornikhansen(): Doornik-Hansen test
jarquebera(): Jarque-Bera test
skewness(): Skewness test
kurtosis(): Kurtosis test
x <- rnorm(100) # null bowmanshenton(x) doornikhansen(x) jarquebera(x) x <- random_t(2, 100) # alternative bowmanshenton(x) doornikhansen(x) jarquebera(x)x <- rnorm(100) # null bowmanshenton(x) doornikhansen(x) jarquebera(x) x <- random_t(2, 100) # alternative bowmanshenton(x) doornikhansen(x) jarquebera(x)
Generating Outlier regressors
ao_variable(frequency, start, length, s, pos, date = NULL) tc_variable(frequency, start, length, s, pos, date = NULL, rate = 0.7) ls_variable(frequency, start, length, s, pos, date = NULL, zeroended = TRUE) so_variable(frequency, start, length, s, pos, date = NULL, zeroended = TRUE)ao_variable(frequency, start, length, s, pos, date = NULL) tc_variable(frequency, start, length, s, pos, date = NULL, rate = 0.7) ls_variable(frequency, start, length, s, pos, date = NULL, zeroended = TRUE) so_variable(frequency, start, length, s, pos, date = NULL, zeroended = TRUE)
frequency |
Frequency of the series, number of periods per year (12,4,3,2..) |
start, length
|
First date (array with the first year and the first period)
(for instance |
s |
time series used to get the dates for the trading days variables. If supplied the
parameters |
pos, date
|
the date of the outlier, defined by the position in period compared to the first date ( |
rate |
the decay rate of the transitory change regressor (see details). |
zeroended |
Boolean indicating if the regressor should end by 0 ( |
An additive outlier (AO, ao_variable) is defined as:
A level shift (LS, ls_variable) is defined as (if zeroended = TRUE):
A transitory change (TC, tc_variable) is defined as:
A seasonal outlier (SO, so_variable) is defined as (if zeroended = TRUE):
# Outliers in February 2002 ao <- ao_variable(12, c(2000, 1), length = 12 * 4, date = "2002-02-01") ls <- ls_variable(12, c(2000, 1), length = 12 * 4, date = "2002-02-01") tc <- tc_variable(12, c(2000, 1), length = 12 * 4, date = "2002-02-01") so <- so_variable(12, c(2000, 1), length = 12 * 4, date = "2002-02-01") plot.ts(ts.union(ao, ls, tc, so), plot.type = "single", col = c("black", "orange", "green", "gray") )# Outliers in February 2002 ao <- ao_variable(12, c(2000, 1), length = 12 * 4, date = "2002-02-01") ls <- ls_variable(12, c(2000, 1), length = 12 * 4, date = "2002-02-01") tc <- tc_variable(12, c(2000, 1), length = 12 * 4, date = "2002-02-01") so <- so_variable(12, c(2000, 1), length = 12 * 4, date = "2002-02-01") plot.ts(ts.union(ao, ls, tc, so), plot.type = "single", col = c("black", "orange", "green", "gray") )
Period splines
periodic_splines(order = 4, period = 1, knots, pos)periodic_splines(order = 4, period = 1, knots, pos)
order |
Order of the splines (4 for cubic) |
period |
Period of the splines (1 by default) |
knots |
Knots of the splines (in [0, period[]]) |
pos |
Requested positions (in [0, period[]]) |
A matrix (len(pos) x len(knots))
Periodic dummies and contrasts
periodic.dummies(frequency, start, length, s) periodic.contrasts(frequency, start, length, s)periodic.dummies(frequency, start, length, s) periodic.contrasts(frequency, start, length, s)
frequency |
Frequency of the series, number of periods per year (12,4,3,2..) |
start, length
|
First date (array with the first year and the first period)
(for instance |
s |
time series used to get the dates for the trading days variables. If supplied the
parameters |
The function periodic.dummies creates as many time series as types of periods in a year (4 or 12) with the value one only for one given type of period (ex Q1) The periodic.contrasts function is based on periodic.dummies but adds -1 to the period preceding a 1.
# periodic dummies for a quarterly series p <- periodic.dummies(4, c(2000, 1), 60) # periodic contrasts for a quarterly series q <- periodic.contrasts(4, c(2000, 1), 60) q[1:9, ]# periodic dummies for a quarterly series p <- periodic.dummies(4, c(2000, 1), 60) # periodic contrasts for a quarterly series q <- periodic.contrasts(4, c(2000, 1), 60) q[1:9, ]
Print functions for calendars
## S3 method for class 'JD3_FIXEDDAY' print(x, ...) ## S3 method for class 'JD3_FIXEDWEEKDAY' print(x, ...) ## S3 method for class 'JD3_EASTERDAY' print(x, ...) ## S3 method for class 'JD3_SPECIALDAY' print(x, ...) ## S3 method for class 'JD3_SINGLEDAY' print(x, ...) ## S3 method for class 'JD3_CALENDAR' print(x, ...)## S3 method for class 'JD3_FIXEDDAY' print(x, ...) ## S3 method for class 'JD3_FIXEDWEEKDAY' print(x, ...) ## S3 method for class 'JD3_EASTERDAY' print(x, ...) ## S3 method for class 'JD3_SPECIALDAY' print(x, ...) ## S3 method for class 'JD3_SINGLEDAY' print(x, ...) ## S3 method for class 'JD3_CALENDAR' print(x, ...)
x |
The object. |
... |
other unused parameters. |
Create Java CalendarTimeSeries
r2jd_calendarts(calendarobs)r2jd_calendarts(calendarobs)
calendarobs |
list. |
obs <- list( list(start = as.Date("1980-01-01"), end = as.Date("1999-12-31"), value = 2000), list(start = as.Date("2000-01-01"), end = as.Date("2010-01-01"), value = 1000) ) jobj <- r2jd_calendarts(obs)obs <- list( list(start = as.Date("1980-01-01"), end = as.Date("1999-12-31"), value = 2000), list(start = as.Date("2000-01-01"), end = as.Date("2010-01-01"), value = 1000) ) jobj <- r2jd_calendarts(obs)
Ramp regressor
ramp_variable(frequency, start, length, s, range)ramp_variable(frequency, start, length, s, range)
frequency |
Frequency of the series, number of periods per year (12,4,3,2..) |
start, length
|
First date (array with the first year and the first period)
(for instance |
s |
time series used to get the dates for the trading days variables. If supplied the
parameters |
range |
the range of the regressor. A vector of length 2 containing the datesin the format |
A ramp between two dates and is defined as:
# Ramp variable from January 2001 to September 2001 rp <- ramp_variable(12, c(2000, 1), length = 12 * 4, range = c(13, 21)) # Or equivalently rp <- ramp_variable(12, c(2000, 1), length = 12 * 4, range = c("2001-01-01", "2001-09-02")) plot.ts(rp)# Ramp variable from January 2001 to September 2001 rp <- ramp_variable(12, c(2000, 1), length = 12 * 4, range = c(13, 21)) # Or equivalently rp <- ramp_variable(12, c(2000, 1), length = 12 * 4, range = c("2001-01-01", "2001-09-02")) plot.ts(rp)
Function to perform a range-mean regression, trimmed to avoid outlier distortion. The can be used to select whether the original series will be transformed into log or maintain in level.
rangemean_tstat(data, period = 0, groupsize = 0, trim = 0)rangemean_tstat(data, period = 0, groupsize = 0, trim = 0)
data |
data to test. |
period |
periodicity of the data. |
groupsize |
number of observations per group (before being trimmed).
The default group size (
|
trim |
number of trimmed observations. |
First, the data is divided into groups of successive observations of length (groupsize).
That is, the first group is formed with the first observations,
the second group is formed with observations to , etc.
Then, for each group , the observations are sorted and the trim smallest and largest
observations are rejected (to avoid outlier distortion).
With the other observations, the range (noted ) and mean (noted ) are computed.
Finally, the following regression is performed :
The function rangemean_tstat returns the T-statistic associated to .
If it is significantly higher than 0, log transformation is recommended.
T-Stat of the slope of the range-mean regression.
y <- ABS$X0.2.09.10.M # Multiplicative pattern plot(y) period <- 12 rm_t <- rangemean_tstat(y, period = period, groupsize = period) rm_t # higher than 0 # Can be tested: pt(rm_t, period - 2, lower.tail = FALSE) # Or : 1 - cdf_t(period - 2, rm_t) # Close to 0 rm_t_log <- rangemean_tstat(log(y), period = period, groupsize = period) rm_t_log pt(rm_t_log, period - 2, lower.tail = FALSE)y <- ABS$X0.2.09.10.M # Multiplicative pattern plot(y) period <- 12 rm_t <- rangemean_tstat(y, period = period, groupsize = period) rm_t # higher than 0 # Can be tested: pt(rm_t, period - 2, lower.tail = FALSE) # Or : 1 - cdf_t(period - 2, rm_t) # Close to 0 rm_t_log <- rangemean_tstat(log(y), period = period, groupsize = period) rm_t_log pt(rm_t_log, period - 2, lower.tail = FALSE)
US Retail trade statistics
retailretail
An object of class list of length 62.
US-Census Bureau
Functions to compute runs test around the mean or the median (testofruns) or up and down runs test (testofupdownruns) to check randomness of a data.
testofruns(data, mean = TRUE, number = TRUE) testofupdownruns(data, number = TRUE)testofruns(data, mean = TRUE, number = TRUE) testofupdownruns(data, number = TRUE)
data |
data being tested. |
mean |
If |
number |
If |
A c("JD3_TEST", "JD3") object (see statisticaltest() for details).
testofruns(): Runs test around mean or median
testofupdownruns(): up and down runs test
x <- random_t(5, 1000) # random values testofruns(x) testofupdownruns(x) # non-random values testofruns(ABS$X0.2.09.10.M) testofupdownruns(ABS$X0.2.09.10.M)x <- random_t(5, 1000) # random values testofruns(x) testofupdownruns(x) # non-random values testofruns(ABS$X0.2.09.10.M) testofupdownruns(ABS$X0.2.09.10.M)
Generic function for preprocessing defined in other packages.
sa_preprocessing(x, ...)sa_preprocessing(x, ...)
x, ...
|
parameters. |
Generic function to format the seasonal adjustment decomposition components.
sa_decomposition() is a generic function defined in other packages.
sadecomposition(y, sa, t, s, i, mul) ## S3 method for class 'JD3_SADECOMPOSITION' print(x, n_last_obs = frequency(x$series), ...) ## S3 method for class 'JD3_SADECOMPOSITION' plot( x, first_date = NULL, last_date = NULL, type_chart = c("sa-trend", "seas-irr"), caption = c(`sa-trend` = "Y, Sa, trend", `seas-irr` = "Sea., irr.")[type_chart], colors = c(y = "#F0B400", t = "#1E6C0B", sa = "#155692", s = "#1E6C0B", i = "#155692"), ... ) sa_decomposition(x, ...)sadecomposition(y, sa, t, s, i, mul) ## S3 method for class 'JD3_SADECOMPOSITION' print(x, n_last_obs = frequency(x$series), ...) ## S3 method for class 'JD3_SADECOMPOSITION' plot( x, first_date = NULL, last_date = NULL, type_chart = c("sa-trend", "seas-irr"), caption = c(`sa-trend` = "Y, Sa, trend", `seas-irr` = "Sea., irr.")[type_chart], colors = c(y = "#F0B400", t = "#1E6C0B", sa = "#155692", s = "#1E6C0B", i = "#155692"), ... ) sa_decomposition(x, ...)
y, sa, t, s, i, mul
|
seasonal adjustment decomposition parameters. |
x |
the object to print. |
n_last_obs |
number of observations to print (by default equal to the frequency of the series). |
... |
further arguments. |
first_date, last_date
|
first and last date to plot (by default all the data is used). |
type_chart |
the chart to plot: |
caption |
the caption of the plot. |
colors |
the colours used in the plot. |
"JD3_SADECOMPOSITION" object.
Decompose SARIMA Model into three components trend, seasonal, irregular
sarima_decompose(model, rmod = 0, epsphi = 0)sarima_decompose(model, rmod = 0, epsphi = 0)
model |
SARIMA model to decompose. |
rmod |
trend threshold. |
epsphi |
seasonal tolerance (in degrees). |
An UCARIMA model
model <- sarima_model(period = 12, d = 1, bd = 1, theta = -0.6, btheta = -0.5) ucm <- sarima_decompose(model)model <- sarima_model(period = 12, d = 1, bd = 1, theta = -0.6, btheta = -0.5) ucm <- sarima_decompose(model)
Estimate SARIMA Model
sarima_estimate( x, order = c(0, 0, 0), seasonal = list(order = c(0, 0, 0), period = NA), mean = FALSE, xreg = NULL, eps = 1e-09 )sarima_estimate( x, order = c(0, 0, 0), seasonal = list(order = c(0, 0, 0), period = NA), mean = FALSE, xreg = NULL, eps = 1e-09 )
x |
a univariate time series. |
order |
vector specifying of the non-seasonal part of the ARIMA model: the AR order, the degree of differencing, and the MA order. |
seasonal |
specification of the seasonal part of the ARIMA model and the seasonal frequency (by default equals to |
mean |
should the SARIMA model include an intercept term. |
xreg |
vector or matrix of external regressors. |
eps |
precision. |
y <- ABS$X0.2.09.10.M sarima_estimate(y, order = c(0, 1, 1), seasonal = c(0, 1, 1))y <- ABS$X0.2.09.10.M sarima_estimate(y, order = c(0, 1, 1), seasonal = c(0, 1, 1))
Title
sarima_hannan_rissanen( x, order = c(0, 0, 0), seasonal = list(order = c(0, 0, 0), period = NA), initialization = c("Ols", "Levinson", "Burg"), biasCorrection = TRUE, finalCorrection = TRUE )sarima_hannan_rissanen( x, order = c(0, 0, 0), seasonal = list(order = c(0, 0, 0), period = NA), initialization = c("Ols", "Levinson", "Burg"), biasCorrection = TRUE, finalCorrection = TRUE )
x |
a univariate time series. |
order |
vector specifying of the non-seasonal part of the ARIMA model: the AR order, the degree of differencing, and the MA order. |
seasonal |
specification of the seasonal part of the ARIMA model and the seasonal frequency (by default equals to |
initialization |
Algorithm used in the computation of the long order auto-regressive model (used to estimate the innovations) |
biasCorrection |
Bias correction |
finalCorrection |
Final correction as implemented in Tramo |
y <- ABS$X0.2.09.10.M sarima_hannan_rissanen(y, order = c(0, 1, 1), seasonal = c(0, 1, 1))y <- ABS$X0.2.09.10.M sarima_hannan_rissanen(y, order = c(0, 1, 1), seasonal = c(0, 1, 1))
Seasonal ARIMA model (Box-Jenkins)
sarima_model( name = "sarima", period, phi = NULL, d = 0, theta = NULL, bphi = NULL, bd = 0, btheta = NULL )sarima_model( name = "sarima", period, phi = NULL, d = 0, theta = NULL, bphi = NULL, bd = 0, btheta = NULL )
name |
name of the model. |
period |
period of the model. |
phi |
coefficients of the regular auto-regressive polynomial
( |
d |
regular differencing order. |
theta |
coefficients of the regular moving average polynomial
( |
bphi |
coefficients of the seasonal auto-regressive polynomial. True signs. |
bd |
seasonal differencing order. |
btheta |
coefficients of the seasonal moving average polynomial. True signs. |
A "JD3_SARIMA" model.
SARIMA Properties
sarima_properties(model, nspectrum = 601, nacf = 36)sarima_properties(model, nspectrum = 601, nacf = 36)
model |
a |
nspectrum |
number of points in [0, pi] to calculate the spectrum. |
nacf |
maximum lag at which to calculate the acf. |
mod1 <- sarima_model(period = 12, d = 1, bd = 1, theta = 0.2, btheta = 0.2) sarima_properties(mod1)mod1 <- sarima_model(period = 12, d = 1, bd = 1, theta = 0.2, btheta = 0.2) sarima_properties(mod1)
Simulate Seasonal ARIMA
sarima_random(model, length, stde = 1, tdegree = 0, seed = -1)sarima_random(model, length, stde = 1, tdegree = 0, seed = -1)
model |
a |
length |
length of the output series. |
stde |
deviation of the normal distribution of the innovations of the simulated series.
Unused if |
tdegree |
degrees of freedom of the T distribution of the innovations.
|
seed |
seed of the random numbers generator. Negative values mean random seeds |
# Airline model s_model <- sarima_model(period = 12, d = 1, bd = 1, theta = 0.2, btheta = 0.2) x <- sarima_random(s_model, length = 64, seed = 0) plot(x, type = "l")# Airline model s_model <- sarima_model(period = 12, d = 1, bd = 1, theta = 0.2, btheta = 0.2) x <- sarima_random(s_model, length = 64, seed = 0) plot(x, type = "l")
Canova-Hansen seasonality test
seasonality_canovahansen( data, period, type = c("Contrast", "Dummy", "Trigonometric"), lag1 = TRUE, kernel = c("Bartlett", "Square", "Welch", "Tukey", "Hamming", "Parzen"), order = NA, start = 1 )seasonality_canovahansen( data, period, type = c("Contrast", "Dummy", "Trigonometric"), lag1 = TRUE, kernel = c("Bartlett", "Square", "Welch", "Tukey", "Hamming", "Parzen"), order = NA, start = 1 )
data |
the input data. |
period |
Tested periodicity. Can be missing if the input is a time series |
type |
Trigonometric variables, seasonal dummies or seasonal contrasts. |
lag1 |
Lagged variable in the regression model. |
kernel |
Kernel used to compute the robust Newey-West covariance matrix. |
order |
The truncation parameter used to compute the robust Newey-West covariance matrix. |
start |
Position of the first observation of the series |
list with the FTest on seasonal variables, the joint test and the details for the stability of the different seasonal variables
s <- log(ABS$X0.2.20.10.M) seasonality_canovahansen(s, 12, type = "Contrast") seasonality_canovahansen(s, 12, type = "Trigonometric")s <- log(ABS$X0.2.20.10.M) seasonality_canovahansen(s, 12, type = "Contrast") seasonality_canovahansen(s, 12, type = "Trigonometric")
Canova-Hansen test using trigonometric variables
seasonality_canovahansen_trigs( data, periods, lag1 = TRUE, kernel = c("Bartlett", "Square", "Welch", "Tukey", "Hamming", "Parzen"), order = NA, original = FALSE )seasonality_canovahansen_trigs( data, periods, lag1 = TRUE, kernel = c("Bartlett", "Square", "Welch", "Tukey", "Hamming", "Parzen"), order = NA, original = FALSE )
data |
the input data. |
periods |
Periodicities. |
lag1 |
Lagged variable in the regression model. |
kernel |
Kernel used to compute the robust Newey-West covariance matrix. |
order |
The truncation parameter used to compute the robust Newey-West covariance matrix. |
original |
|
s <- log(ABS$X0.2.20.10.M) freqs <- seq(0.01, 0.5, 0.001) plot(seasonality_canovahansen_trigs(s, 1 / freqs, original = FALSE), type = "l")s <- log(ABS$X0.2.20.10.M) freqs <- seq(0.01, 0.5, 0.001) plot(seasonality_canovahansen_trigs(s, 1 / freqs, original = FALSE), type = "l")
"X12" Test On Seasonality
seasonality_combined( data, period = NA, firstperiod = cycle(data)[1], mul = TRUE )seasonality_combined( data, period = NA, firstperiod = cycle(data)[1], mul = TRUE )
data |
the input data. |
period |
Tested periodicity. Can be missing if the input is a time series |
firstperiod |
Position in a cycle of the first obs.
For example, for a monthly, |
mul |
boolean indicating if the seasonal decomposition is multiplicative ( |
Combined test on the presence of identifiable seasonality (see Ladiray and Quenneville, 1999).
s <- do_stationary(log(ABS$X0.2.09.10.M))$ddata seasonality_combined(s) seasonality_combined(random_t(2, 1000), 7)s <- do_stationary(log(ABS$X0.2.09.10.M))$ddata seasonality_combined(s) seasonality_combined(random_t(2, 1000), 7)
F-test on seasonal dummies
seasonality_f(data, period = NA, model = c("AR", "D1", "WN"), nyears = 0)seasonality_f(data, period = NA, model = c("AR", "D1", "WN"), nyears = 0)
data |
the input data. |
period |
Tested periodicity. Can be missing if the input is a time series |
model |
the model to use for the residuals. |
nyears |
Number of periods or number of cycles considered in the test, at the end of the series:
in periods (positive value) or years (negative values).
By default ( |
Estimation of a model with seasonal dummies. Joint F-test on the coefficients of the dummies.
A c("JD3_TEST", "JD3") object (see statisticaltest() for details).
seasonality_f(ABS$X0.2.09.10.M, model = "D1") seasonality_f(random_t(2, 1000), 7)seasonality_f(ABS$X0.2.09.10.M, model = "D1") seasonality_f(random_t(2, 1000), 7)
Friedman Seasonality Test
seasonality_friedman(data, period = NA, nyears = 0)seasonality_friedman(data, period = NA, nyears = 0)
data |
the input data. |
period |
Tested periodicity. Can be missing if the input is a time series |
nyears |
Number of periods or number of cycles considered in the test, at the end of the series:
in periods (positive value) or years (negative values).
By default ( |
Non parametric test ("ANOVA"-type).
A c("JD3_TEST", "JD3") object (see statisticaltest() for details).
s <- do_stationary(log(ABS$X0.2.09.10.M))$ddata seasonality_friedman(s) seasonality_friedman(random_t(2, 1000), 12)s <- do_stationary(log(ABS$X0.2.09.10.M))$ddata seasonality_friedman(s) seasonality_friedman(random_t(2, 1000), 12)
Kruskall-Wallis Seasonality Test
seasonality_kruskalwallis(data, period, nyears = 0)seasonality_kruskalwallis(data, period, nyears = 0)
data |
the input data. |
period |
Tested periodicity. Can be missing if the input is a time series |
nyears |
Number of periods or number of cycles considered in the test, at the end of the series:
in periods (positive value) or years (negative values).
By default ( |
Non parametric test on the ranks.
A c("JD3_TEST", "JD3") object (see statisticaltest() for details).
s <- do_stationary(log(ABS$X0.2.09.10.M))$ddata seasonality_kruskalwallis(s) seasonality_kruskalwallis(random_t(2, 1000), 7)s <- do_stationary(log(ABS$X0.2.09.10.M))$ddata seasonality_kruskalwallis(s) seasonality_kruskalwallis(random_t(2, 1000), 7)
Modified QS Seasonality Test (Maravall)
seasonality_modified_qs(data, period = NA, nyears = 0)seasonality_modified_qs(data, period = NA, nyears = 0)
data |
the input data. |
period |
Tested periodicity. Can be missing if the input is a time series |
nyears |
Number of periods or number of cycles considered in the test, at the end of the series:
in periods (positive value) or years (negative values).
By default ( |
Thresholds for p-values: p.9=2.49, p.95=3.83, p.99=7.06, p.999=11.88. Computed on 100.000.000 random series (different lengths). Remark: the length of the series has some impact on the p-values, mainly on short series. Not critical.
The value of the test
s <- do_stationary(log(ABS$X0.2.09.10.M))$ddata seasonality_modified_qs(s)s <- do_stationary(log(ABS$X0.2.09.10.M))$ddata seasonality_modified_qs(s)
Periodogram Seasonality Test
seasonality_periodogram(data, period = NA, nyears = 0)seasonality_periodogram(data, period = NA, nyears = 0)
data |
the input data. |
period |
Tested periodicity. Can be missing if the input is a time series |
nyears |
Number of periods or number of cycles considered in the test, at the end of the series:
in periods (positive value) or years (negative values).
By default ( |
Tests on the sum of a periodogram at seasonal frequencies.
A c("JD3_TEST", "JD3") object (see statisticaltest() for details).
s <- do_stationary(log(ABS$X0.2.09.10.M))$ddata seasonality_periodogram(s) seasonality_periodogram(random_t(2, 1000), 7)s <- do_stationary(log(ABS$X0.2.09.10.M))$ddata seasonality_periodogram(s) seasonality_periodogram(random_t(2, 1000), 7)
QS (seasonal Ljung-Box) test.
seasonality_qs(data, period = NA, nyears = 0, type = 1)seasonality_qs(data, period = NA, nyears = 0, type = 1)
data |
the input data. |
period |
Tested periodicity. Can be missing if the input is a time series |
nyears |
Number of periods or number of cycles considered in the test, at the end of the series:
in periods (positive value) or years (negative values).
By default ( |
type |
1 for positive autocorrelations, -1 for negative autocorrelations,
0 for all autocorrelations. By default ( |
A c("JD3_TEST", "JD3") object (see statisticaltest() for details).
s <- do_stationary(log(ABS$X0.2.09.10.M))$ddata seasonality_qs(s) seasonality_qs(random_t(2, 1000), 7)s <- do_stationary(log(ABS$X0.2.09.10.M))$ddata seasonality_qs(s) seasonality_qs(random_t(2, 1000), 7)
Function allowing to customize the ARIMA model structure when the automatic modelling is disabled.(see example)
set_arima( x, mean = NA, mean.type = c(NA, "Undefined", "Fixed", "Initial"), p = NA, d = NA, q = NA, bp = NA, bd = NA, bq = NA, coef = NA, coef.type = c(NA, "Undefined", "Fixed", "Initial") )set_arima( x, mean = NA, mean.type = c(NA, "Undefined", "Fixed", "Initial"), p = NA, d = NA, q = NA, bp = NA, bd = NA, bq = NA, coef = NA, coef.type = c(NA, "Undefined", "Fixed", "Initial") )
x |
the specification to customize, must be a "SPEC" class object (see details). |
mean |
to fix the coefficient of the mean. If |
mean.type |
a character defining the mean coefficient estimation procedure.
Possible procedures are: |
p, d, q, bp, bd, bq
|
to specify the order of the SARIMA model in the form ARIMA(p,d,q)(bp,bd,bd). |
coef |
a vector providing the coefficients for the regular and seasonal AR and MA polynomials.
The vector length must be equal to the sum of the regular and seasonal AR and MA orders.
The coefficients shall be provided in the following order: regular AR (Phi; |
coef.type |
a vector defining the ARMA coefficients estimation procedure.
Possible procedures are: |
x specification parameter must be a JD3_X13_SPEC" class object generated with rjd3x13::x13_spec()
(or "JD3_REGARIMA_SPEC" generated with rjd3x13::spec_regarima() or "JD3_TRAMOSEATS_SPEC"
generated with rjd3tramoseats::spec_tramoseats() or "JD3_TRAMO_SPEC" generated with
rjd3tramoseats::spec_tramo()).
More information on reg-arima modelling in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/
# create default spec # my_spec<-rjd3x13::x13_spec("rsa5c") # disable automatic arima modelling # my_spec<-set_automodel(my_spec, enabled = FALSE) # customize arima model # my_spec <-set_arima(my_spec,mean = 0.2, # mean.type = "Fixed", # p = 1, d = 2, q = 0, # bp = 1, bd = 1, bq = 0, # coef = c(0.6,0.7), # coef.type = c("Initial","Fixed"))# create default spec # my_spec<-rjd3x13::x13_spec("rsa5c") # disable automatic arima modelling # my_spec<-set_automodel(my_spec, enabled = FALSE) # customize arima model # my_spec <-set_arima(my_spec,mean = 0.2, # mean.type = "Fixed", # p = 1, d = 2, q = 0, # bp = 1, bd = 1, bq = 0, # coef = c(0.6,0.7), # coef.type = c("Initial","Fixed"))
Function allowing to customize Arima model identification procedure.
set_automodel( x, enabled = NA, acceptdefault = NA, cancel = NA, ub1 = NA, ub2 = NA, reducecv = NA, ljungboxlimit = NA, tsig = NA, ubfinal = NA, checkmu = NA, mixed = NA, fct = NA, balanced = NA, amicompare = NA )set_automodel( x, enabled = NA, acceptdefault = NA, cancel = NA, ub1 = NA, ub2 = NA, reducecv = NA, ljungboxlimit = NA, tsig = NA, ubfinal = NA, checkmu = NA, mixed = NA, fct = NA, balanced = NA, amicompare = NA )
x |
the specification to customize, must be a "SPEC" class object (see details). |
enabled |
|
acceptdefault |
|
cancel |
|
ub1 |
|
ub2 |
|
reducecv |
|
ljungboxlimit |
|
tsig |
|
ubfinal |
(REGARIMA/X13 Specific) |
checkmu |
(REGARIMA/X13 Specific) |
mixed |
(REGARIMA/X13 Specific) |
fct |
(REGARIMA/X13 Specific) |
balanced |
(REGARIMA/X13 Specific) |
amicompare |
(TRAMO Specific) |
x specification parameter must be a JD3_X13_SPEC" class object generated with rjd3x13::x13_spec()
(or "JD3_REGARIMA_SPEC" generated with rjd3x13::spec_regarima() or "JD3_TRAMOSEATS_SPEC"
generated with rjd3tramoseats::spec_tramoseats() or "JD3_TRAMO_SPEC" generated with
rjd3tramoseats::spec_tramo()).
More information on reg-arima modelling in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/
# init_spec <- rjd3x13::x13_spec("RSA5c") # new_spec<-set_automodel(init_spec, # enabled = FALSE, # acceptdefault = TRUE)# init_spec <- rjd3x13::x13_spec("RSA5c") # new_spec<-set_automodel(init_spec, # enabled = FALSE, # acceptdefault = TRUE)
Function allowing to check if the series can be processed and to define a sub-span on which estimation will be performed
set_basic( x, type = c(NA, "All", "From", "To", "Between", "Last", "First", "Excluding"), d0 = NULL, d1 = NULL, n0 = 0, n1 = 0, preliminary.check = NA, preprocessing = NA )set_basic( x, type = c(NA, "All", "From", "To", "Between", "Last", "First", "Excluding"), d0 = NULL, d1 = NULL, n0 = 0, n1 = 0, preliminary.check = NA, preprocessing = NA )
x |
the specification to customize, must be a "SPEC" class object (see details). |
type, d0, d1, n0, n1
|
parameters to specify the sub-span .
|
preliminary.check |
a Boolean to check the quality of the input series and exclude highly problematic ones (e.g. the series with a number of identical observations and/or missing values above pre-specified threshold values). |
preprocessing |
(REGARIMA/X13 Specific) a Boolean to enable/disable the pre-processing. Option disabled for the moment. |
x specification parameter must be a JD3_X13_SPEC" class object
generated with rjd3x13::x13_spec() (or "JD3_REGARIMA_SPEC" generated
with rjd3x13::spec_regarima() or "JD3_TRAMOSEATS_SPEC" generated with
rjd3tramoseats::spec_tramoseats() or "JD3_TRAMO_SPEC" generated with
rjd3tramoseats::spec_tramo()).
More information in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/
# init_spec <- rjd3x13::x13_spec("RSA5c") # estimation on sub-span between two dates (date d1 is excluded) # new_spec<-set_basic(init_spec,type = "Between",d0 = "2014-01-01", # d1 = "2019-01-01", preliminary.check = TRUE, preprocessing = TRUE) # Estimation on the first 60 observations # new_spec <-set_basic(init_spec,Type="First", n0 = 60, # preliminary.check = TRUE, # preprocessing= TRUE) # Estimation on the last 60 observations # new_spec <-set_basic(init_spec,Type="Last", n1 = 60, # preliminary.check = TRUE, # preprocessing= TRUE) # Estimation excluding 60 observations at the beginning and 36 at the end of the series # new_spec <-set_basic(init_spec,Type="Excluding", n0=60, n1=36, # preliminary.check = TRUE, # preprocessing= TRUE)# init_spec <- rjd3x13::x13_spec("RSA5c") # estimation on sub-span between two dates (date d1 is excluded) # new_spec<-set_basic(init_spec,type = "Between",d0 = "2014-01-01", # d1 = "2019-01-01", preliminary.check = TRUE, preprocessing = TRUE) # Estimation on the first 60 observations # new_spec <-set_basic(init_spec,Type="First", n0 = 60, # preliminary.check = TRUE, # preprocessing= TRUE) # Estimation on the last 60 observations # new_spec <-set_basic(init_spec,Type="Last", n1 = 60, # preliminary.check = TRUE, # preprocessing= TRUE) # Estimation excluding 60 observations at the beginning and 36 at the end of the series # new_spec <-set_basic(init_spec,Type="Excluding", n0=60, n1=36, # preliminary.check = TRUE, # preprocessing= TRUE)
Function allowing to perform a benchmarking procedure after the decomposition step in a seasonal adjustment (disabled by default). Here benchmarking refers to a procedure ensuring consistency over the year between seasonally adjusted and raw (or calendar adjusted) data, as seasonal adjustment can cause discrepancies between the annual totals of seasonally adjusted series and the corresponding annual totals of raw (or calendar adjusted) series.
set_benchmarking( x, enabled = NA, target = c(NA, "CalendarAdjusted", "Original"), rho = NA, lambda = NA, forecast = NA, bias = c(NA, "None") )set_benchmarking( x, enabled = NA, target = c(NA, "CalendarAdjusted", "Original"), rho = NA, lambda = NA, forecast = NA, bias = c(NA, "None") )
x |
the specification to customize, must be a "SPEC" class object (see details). |
enabled |
Boolean to enable the user to perform benchmarking. |
target |
specifies the target series for the benchmarking procedure,
which can be the raw series ( |
rho |
the value of the AR(1) parameter (set between 0 and 1) in the function used for benchmarking. Default =1. |
lambda |
a parameter in the function used for benchmarking that relates to the weights in the regression equation; it is typically equal to 0, 1/2 or 1. |
forecast |
Boolean indicating if the forecasts of the seasonally
adjusted series and of the target variable ( |
bias |
TODO |
x specification parameter must be a JD3_X13_SPEC" class object
generated with rjd3x13::x13_spec() (or "JD3_REGARIMA_SPEC" generated
with rjd3x13::spec_regarima() or "JD3_TRAMOSEATS_SPEC" generated with
rjd3tramoseats::spec_tramoseats() or "JD3_TRAMO_SPEC" generated with
rjd3tramoseats::spec_tramo()).
More information on benchmarking in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/
# init_spec <- rjd3x13::x13_spec("RSA5c") # new_spec<- set_benchmarking(init_spec, # enabled = TRUE, # target = "Normal", # rho = 0.8, # lambda = 0.5, # forecast = FALSE, # bias = "None")# init_spec <- rjd3x13::x13_spec("RSA5c") # new_spec<- set_benchmarking(init_spec, # enabled = TRUE, # target = "Normal", # rho = 0.8, # lambda = 0.5, # forecast = FALSE, # bias = "None")
Set Easter effect correction in Pre-Processing Specification
set_easter( x, enabled = NA, julian = NA, duration = NA, test = c(NA, "Add", "Remove", "None"), coef = NA, coef.type = c(NA, "Estimated", "Fixed"), type = c(NA, "Unused", "Standard", "IncludeEaster", "IncludeEasterMonday") )set_easter( x, enabled = NA, julian = NA, duration = NA, test = c(NA, "Add", "Remove", "None"), coef = NA, coef.type = c(NA, "Estimated", "Fixed"), type = c(NA, "Unused", "Standard", "IncludeEaster", "IncludeEasterMonday") )
x |
the specification to customize, must be a "SPEC" class object (see details). |
enabled |
a logical indicating if the program considers the Easter effect in the pre-processing model. Default = TRUE. |
julian |
a logical indicating if the program uses the Julian Easter (expressed in Gregorian calendar). |
duration |
a numeric indicating the duration of the Easter effect (length in days, between 1 and 20). Default value = 8 in REGARIMA/X-13 and 6 in TRAMO. |
test |
defines the pre-tests for the significance of the Easter effect based on the t-statistic
(the Easter effect is considered as significant if the t-statistic is greater than 1.96):
|
coef |
to set the coefficient of the easter regressor.(Test parameter has to be set to |
coef.type |
a character defining the easter regressor coefficient estimation procedure.
Possible procedures are: |
type |
(TRAMO specific) a |
x specification parameter must be a JD3_X13_SPEC" class object generated with rjd3x13::x13_spec()
(or "JD3_REGARIMA_SPEC" generated with rjd3x13::spec_regarima() or "JD3_TRAMOSEATS_SPEC"
generated with rjd3tramoseats::spec_tramoseats() or "JD3_TRAMO_SPEC" generated with
rjd3tramoseats::spec_tramo()).
More information on calendar correction in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/a-calendar-correction
# init_spec <- rjd3x13::x13_spec("RSA5c") # new_spec<-set_easter(init_spec, # enabled = TRUE, # duration = 12, # test = "None", # type = "IncludeEasterMonday") # sa<-rjd3x13::x13(ABS$X0.2.09.10.M,new_spec)# init_spec <- rjd3x13::x13_spec("RSA5c") # new_spec<-set_easter(init_spec, # enabled = TRUE, # duration = 12, # test = "None", # type = "IncludeEasterMonday") # sa<-rjd3x13::x13(ABS$X0.2.09.10.M,new_spec)
Function allowing to define numeric boundaries for estimation and to define a sub-span on which reg-arima (tramo) modelling will be performed (pre-processing step)
set_estimate( x, type = c(NA, "All", "From", "To", "Between", "Last", "First", "Excluding"), d0 = NULL, d1 = NULL, n0 = 0, n1 = 0, tol = NA, exact.ml = NA, unit.root.limit = NA )set_estimate( x, type = c(NA, "All", "From", "To", "Between", "Last", "First", "Excluding"), d0 = NULL, d1 = NULL, n0 = 0, n1 = 0, tol = NA, exact.ml = NA, unit.root.limit = NA )
x |
the specification to customize, must be a "SPEC" class object (see details). |
type, d0, d1, n0, n1
|
parameters to specify the sub-span .
|
tol |
a numeric, convergence tolerance. The absolute changes in the log-likelihood function are compared to this value to check for the convergence of the estimation iterations. (The default setting is 0.0000001) |
exact.ml |
(TRAMO specific) |
unit.root.limit |
(TRAMO specific) |
x specification parameter must be a JD3_X13_SPEC" class object
generated with rjd3x13::x13_spec() (or "JD3_REGARIMA_SPEC" generated
with rjd3x13::spec_regarima() or "JD3_TRAMOSEATS_SPEC" generated with
rjd3tramoseats::spec_tramoseats() or "JD3_TRAMO_SPEC" generated with
rjd3tramoseats::spec_tramo()).
More in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/
# init_spec <- rjd3tramoseats::spec_tramoseats("rsafull") # new_spec<-set_estimate(init_spec, type= "From", d0 = "2012-01-01", tol = 0.0000002, # exact.ml = FALSE, unit.root.limit = 0.98)# init_spec <- rjd3tramoseats::spec_tramoseats("rsafull") # new_spec<-set_estimate(init_spec, type= "From", d0 = "2012-01-01", tol = 0.0000002, # exact.ml = FALSE, unit.root.limit = 0.98)
Function allowing to customize the automatic outlier detection process built in in the pre-processing step (regarima or tramo)
set_outlier( x, span.type = c(NA, "All", "From", "To", "Between", "Last", "First", "Excluding"), d0 = NULL, d1 = NULL, n0 = 0, n1 = 0, outliers.type = NA, critical.value = NA, tc.rate = NA, method = c(NA, "AddOne", "AddAll"), maxiter = NA, lsrun = NA, eml.est = NA )set_outlier( x, span.type = c(NA, "All", "From", "To", "Between", "Last", "First", "Excluding"), d0 = NULL, d1 = NULL, n0 = 0, n1 = 0, outliers.type = NA, critical.value = NA, tc.rate = NA, method = c(NA, "AddOne", "AddAll"), maxiter = NA, lsrun = NA, eml.est = NA )
x |
the specification to customize, must be a "SPEC" class object (see details). |
span.type, d0, d1, n0, n1
|
parameters to specify the sub-span on which outliers will be detected.
|
outliers.type |
vector of characters of the outliers to be automatically detected. |
critical.value |
|
tc.rate |
the rate of decay for the transitory change outlier (Default = 0.7). |
method |
(REGARIMA/X13 Specific) determines how the program successively adds detected outliers to the model.
Currently, only the |
maxiter |
(REGARIMA/X13 Specific) maximum number of iterations (Default = 30). |
lsrun |
(REGARIMA/X13 Specific) number of successive level shifts to test for cancellation (Default = 0). |
eml.est |
(TRAMO Specific) |
x specification parameter must be a JD3_X13_SPEC" class object generated with rjd3x13::x13_spec()
(or "JD3_REGARIMA_SPEC" generated with rjd3x13::spec_regarima() or "JD3_TRAMOSEATS_SPEC"
generated with rjd3tramoseats::spec_tramoseats() or "JD3_TRAMO_SPEC" generated with
rjd3tramoseats::spec_tramo()).
If a Seasonal adjustment process is performed, each type of Outlier will be allocated to a pre-defined component after the decomposition: "AO" and "TC" to the irregular, "LS" to the trend and "SO" to seasonal component.
More information on outliers and other auxiliary variables in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/
# init_spec <- rjd3tramoseats::spec_tramoseats("rsafull") # new_spec<-set_outlier(init_spec, span.type= "From", d0 = "2012-01-01", # outliers.type = c("LS", "AO"), # critical.value = 5, # tc.rate =0.85)# init_spec <- rjd3tramoseats::spec_tramoseats("rsafull") # new_spec<-set_outlier(init_spec, span.type= "From", d0 = "2012-01-01", # outliers.type = c("LS", "AO"), # critical.value = 5, # tc.rate =0.85)
Function allowing to select the trading-days regressors to be used for
calendar correction in the pre-processing step of a seasonal adjustment
procedure. The default is "TradingDays", with easter specific effect
enabled. (see set_easter)
All the built-in regressors are meant to correct for type of day effect but don't take into account any holiday. To do so user-defined regressors have to be built.
set_tradingdays( x, option = c(NA, "TradingDays", "WorkingDays", "TD3", "TD3c", "TD4", "None", "UserDefined"), calendar.name = NA, uservariable = NA, stocktd = NA, test = c(NA, "None", "Remove", "Add", "Separate_T", "Joint_F"), coef = NA, coef.type = c(NA, "Fixed", "Estimated"), automatic = c(NA, "Unused", "FTest", "WaldTest", "Aic", "Bic"), pftd = NA, autoadjust = NA, leapyear = c(NA, "LeapYear", "LengthOfPeriod", "None"), leapyear.coef = NA, leapyear.coef.type = c(NA, "Fixed", "Estimated") )set_tradingdays( x, option = c(NA, "TradingDays", "WorkingDays", "TD3", "TD3c", "TD4", "None", "UserDefined"), calendar.name = NA, uservariable = NA, stocktd = NA, test = c(NA, "None", "Remove", "Add", "Separate_T", "Joint_F"), coef = NA, coef.type = c(NA, "Fixed", "Estimated"), automatic = c(NA, "Unused", "FTest", "WaldTest", "Aic", "Bic"), pftd = NA, autoadjust = NA, leapyear = c(NA, "LeapYear", "LengthOfPeriod", "None"), leapyear.coef = NA, leapyear.coef.type = c(NA, "Fixed", "Estimated") )
x |
the specification to customize, must be a "SPEC" class object (see details). |
option |
to specify the set of trading days regression variables:
|
calendar.name |
name (string) of the user-defined calendar to be taken into account when generating built-in regressors set in 'option' (if not 'UserDefined).(see examples) |
uservariable |
a vector of characters to specify the name of user-defined calendar regressors.
When specified, automatically set |
stocktd |
a numeric indicating the day of the month when inventories and other stock are reported
(to denote the last day of the month, set the variable to 31).
When specified, automatically set |
test |
defines the pre-tests for the significance of the trading day regression variables
based on the AICC statistics: (REGARIMA/X-13 specific)
(TRAMO specific)
|
coef |
vector of coefficients for the trading-days regressors. |
coef.type, leapyear.coef.type
|
vector defining if the coefficients are fixed or estimated. |
automatic |
defines whether the calendar effects should be added to the
model manually ( |
pftd |
(TRAMO SPECIFIC) |
autoadjust |
a logical indicating if the program corrects automatically the raw series for the leap year effect if the leap year regressor is significant. Only used when the data is log transformed. |
leapyear |
a |
leapyear.coef |
coefficient of the leap year regressor. |
x specification parameter must be a JD3_X13_SPEC" class object generated with rjd3x13::x13_spec()
(or "JD3_REGARIMA_SPEC" generated with rjd3x13::spec_regarima() or "JD3_TRAMOSEATS_SPEC"
generated with rjd3tramoseats::spec_tramoseats() or "JD3_TRAMO_SPEC" generated with
rjd3tramoseats::spec_tramo()).
More information on calendar correction in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/a-calendar-correction
modelling_context, calendar_td
# Pre-defined regressors # y_raw<-ABS$X0.2.09.10.M # init_spec <- rjd3x13::x13_spec("RSA5c") # new_spec<-set_tradingdays(init_spec, # option = "TD4", # test = "None", # coef=c(0.7,NA,0.5), # coef.type=c("Fixed","Estimated","Fixed"), # leapyear="LengthOfPeriod", # leapyear.coef=0.6) # sa<-rjd3x13::x13(y_raw,new_spec) # Pre-defined regressors based on user-defined calendar ### create a calendar BE <- national_calendar(list( fixed_day(7, 21), special_day("NEWYEAR"), special_day("CHRISTMAS"), special_day("MAYDAY"), special_day("EASTERMONDAY"), special_day("ASCENSION"), special_day("WHITMONDAY"), special_day("ASSUMPTION"), special_day("ALLSAINTSDAY"), special_day("ARMISTICE") )) ## put into a context my_context <- modelling_context(calendars = list(cal = BE)) ## create a specification # init_spec <- rjd3x13::x13_spec("RSA5c") ## modify the specification # new_spec<-set_tradingdays(init_spec, # option = "TradingDays", calendar.name="cal") ## estimate with context # sa<-rjd3x13::x13(y_raw,new_spec, context=my_context) # User-defined regressors # init_spec <- rjd3x13::x13_spec("RSA5c") # add regressors to context # variables<-list(Monday,Tuesday, Wednesday, # Thursday, Friday, Saturday) # my_context<-modelling_context(variables=variables) # create a new spec (here default group name: r) # new_spec<-set_tradingdays(init_spec, # option = "UserDefined", # uservariable=c("r.Monday","r.Tuesday","r.Wednesday","r.Thursday","r.Friday","r.Saturday"), # test = "None") # estimate with context # sa<-rjd3x13::x13(y_raw,new_spec, context=my_context)# Pre-defined regressors # y_raw<-ABS$X0.2.09.10.M # init_spec <- rjd3x13::x13_spec("RSA5c") # new_spec<-set_tradingdays(init_spec, # option = "TD4", # test = "None", # coef=c(0.7,NA,0.5), # coef.type=c("Fixed","Estimated","Fixed"), # leapyear="LengthOfPeriod", # leapyear.coef=0.6) # sa<-rjd3x13::x13(y_raw,new_spec) # Pre-defined regressors based on user-defined calendar ### create a calendar BE <- national_calendar(list( fixed_day(7, 21), special_day("NEWYEAR"), special_day("CHRISTMAS"), special_day("MAYDAY"), special_day("EASTERMONDAY"), special_day("ASCENSION"), special_day("WHITMONDAY"), special_day("ASSUMPTION"), special_day("ALLSAINTSDAY"), special_day("ARMISTICE") )) ## put into a context my_context <- modelling_context(calendars = list(cal = BE)) ## create a specification # init_spec <- rjd3x13::x13_spec("RSA5c") ## modify the specification # new_spec<-set_tradingdays(init_spec, # option = "TradingDays", calendar.name="cal") ## estimate with context # sa<-rjd3x13::x13(y_raw,new_spec, context=my_context) # User-defined regressors # init_spec <- rjd3x13::x13_spec("RSA5c") # add regressors to context # variables<-list(Monday,Tuesday, Wednesday, # Thursday, Friday, Saturday) # my_context<-modelling_context(variables=variables) # create a new spec (here default group name: r) # new_spec<-set_tradingdays(init_spec, # option = "UserDefined", # uservariable=c("r.Monday","r.Tuesday","r.Wednesday","r.Thursday","r.Friday","r.Saturday"), # test = "None") # estimate with context # sa<-rjd3x13::x13(y_raw,new_spec, context=my_context)
Set Log-level Transformation and Decomposition scheme in Pre-Processing Specification
set_transform( x, fun = c(NA, "Auto", "Log", "None"), adjust = c(NA, "None", "LeapYear", "LengthOfPeriod"), outliers = NA, aicdiff = NA, fct = NA )set_transform( x, fun = c(NA, "Auto", "Log", "None"), adjust = c(NA, "None", "LeapYear", "LengthOfPeriod"), outliers = NA, aicdiff = NA, fct = NA )
x |
the specification to customize, must be a "SPEC" class object (see details). |
fun |
the transformation of the input series: |
adjust |
pre-adjustment of the input series for the length of period or leap year effects:
|
outliers |
Boolean indicating if a pre-correction for large outliers (AO and LS only) should be done
in the test for the log-level specification ( |
aicdiff |
(REGARIMA/X-13 specific) a numeric defining the difference in AICC needed to accept no transformation when the automatic
transformation selection is chosen (considered only when |
fct |
(TRAMO specific) |
x specification parameter must be a JD3_X13_SPEC" class object generated with rjd3x13::x13_spec()
(or "JD3_REGARIMA_SPEC" generated with rjd3x13::spec_regarima() or "JD3_TRAMOSEATS_SPEC"
generated with rjd3tramoseats::spec_tramoseats() or "JD3_TRAMO_SPEC" generated with
rjd3tramoseats::spec_tramo()).
More information in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/
# init_spec <- rjd3x13::x13_spec("RSA5c") # new_spec<- set_transform(init_spec, # fun = "Log", # outliers = TRUE) # sa<-rjd3x13::x13(ABS$X0.2.09.10.M,new_spec)# init_spec <- rjd3x13::x13_spec("RSA5c") # new_spec<- set_transform(init_spec, # fun = "Log", # outliers = TRUE) # sa<-rjd3x13::x13(ABS$X0.2.09.10.M,new_spec)
Allows to set a holiday as a once-occurring event.
single_day(date, weight = 1)single_day(date, weight = 1)
date |
the date of the holiday in the format |
weight |
weight associated to the holiday. |
More information on calendar correction in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/a-calendar-correction
national_calendar, fixed_day,special_day,easter_day
single_day("1999-03-19")single_day("1999-03-19")
Allows to define a holiday choosing from a list of pre-specified events, equivalent
to use fixed_day or easter_day functions.
special_day(event, offset = 0, weight = 1, validity = NULL)special_day(event, offset = 0, weight = 1, validity = NULL)
event |
the event to add (see details). |
offset |
The position of the holiday in relation to the selected pre-specified holiday measured in days (can be positive or negative).
By default |
weight |
weight associated to the holiday. |
validity |
validity period: either |
Possible values :
| NEWYEAR | Fixed holiday, falls on January, 1st. |
| SHROVEMONDAY | Moving holiday, falls on the Monday before Ash Wednesday (48 days before Easter Sunday). |
| SHROVETUESDAY | Moving holiday, falls on the Tuesday before Ash Wednesday (47 days before Easter Sunday). |
| ASHWEDNESDAY | Moving holiday, occurring 46 days before Easter Sunday. |
| MAUNDYTHURSDAY | Moving holiday, falls on the Thursday before Easter. |
| GOODFRIDAY | Moving holiday, falls on the Friday before Easter. |
| EASTER | Moving holiday, falls between March 22nd and April 25th. |
| EASTERMONDAY | Moving holiday, falls on the day after Easter. |
| ASCENSION | Moving holiday, celebrated on a Thursday, 39 days after Easter. |
| PENTECOST | Moving holiday, celebrated 49 days after Easter Sunday. |
| WHITMONDAY | Moving holiday, falling on the day after Pentecost. |
| CORPUSCHRISTI | Moving holiday, celebrated 60 days after Easter Sunday. |
| JULIANEASTER | |
| MAYDAY | Fixed holiday, falls on May, 1st. |
| ASSUMPTION | Fixed holiday, falls on August, 15th. |
| HALLOWEEN | Fixed holiday, falls on October, 31st. |
| ALLSAINTSDAY | Fixed holiday, falls on November, 1st. |
| ARMISTICE | Fixed holiday, falls on November, 11th. |
| CHRISTMAS | Fixed holiday, falls on December, 25th. |
More information on calendar correction in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/a-calendar-correction
national_calendar, fixed_day, easter_day
# To add Easter Monday special_day("EASTERMONDAY") # To define a holiday for the day after Christmas, with validity and weight special_day("CHRISTMAS", offset = 1, weight = 0.8, validity = list(start = "2000-01-01", end = "2020-12-01") )# To add Easter Monday special_day("EASTERMONDAY") # To define a holiday for the day after Christmas, with validity and weight special_day("CHRISTMAS", offset = 1, weight = 0.8, validity = list(start = "2000-01-01", end = "2020-12-01") )
Generic function to format the results of 'JDemetra+' tests.
statisticaltest(val, pval, dist = NULL) ## S3 method for class 'JD3_TEST' print(x, details = FALSE, ...)statisticaltest(val, pval, dist = NULL) ## S3 method for class 'JD3_TEST' print(x, details = FALSE, ...)
val, pval, dist
|
statistical parameters. |
x |
the object to print. |
details |
boolean indicating if the statistical distribution should be printed. |
... |
further arguments (ignored). |
c("JD3_TEST", "JD3") object that is a list of three parameters:
value the statistical value of the test.
pvalue the p-value of the test.
distribution the statistical distribution used.
udr_test <- testofupdownruns(random_t(5, 1000)) udr_test # default print print(udr_test, details = TRUE) # with the distributionudr_test <- testofupdownruns(random_t(5, 1000)) udr_test # default print print(udr_test, details = TRUE) # with the distribution
Allows to generate a specific regressor for correcting trading days effects in Stock series.
stock_td(frequency, start, length, s, w = 31)stock_td(frequency, start, length, s, w = 31)
frequency |
Frequency of the series, number of periods per year (12,4,3,2..) |
start, length
|
First date (array with the first year and the first period)
(for instance |
s |
time series used to get the dates for the trading days variables. If supplied the
parameters |
w |
indicates day of the month when inventories and other stocks are reported. (to denote the last day of the month enter 31). |
The regressor will have the value -1 if the w-th day is a Sunday, 1 if it is a Monday as 0 otherwise.
Time series (object of class c("ts","mts","matrix")).
More information on calendar correction in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/a-calendar-correction
Allows to generate trading day regressors (as many as defined groups), taking into account 7 or less different types of days, from Monday to Sunday, but no specific holidays. Regressors are not corrected for long term mean.
td( frequency, start, length, s, groups = c(1, 2, 3, 4, 5, 6, 0), contrasts = TRUE )td( frequency, start, length, s, groups = c(1, 2, 3, 4, 5, 6, 0), contrasts = TRUE )
frequency |
Frequency of the series, number of periods per year (12,4,3,2..) |
start, length
|
First date (array with the first year and the first period)
(for instance |
s |
time series used to get the dates for the trading days variables. If supplied the
parameters |
groups |
Groups of days. The length of the array must be 7. It indicates to what group each week day belongs. The first item corresponds to Mondays and the last one to Sundays. The group used for contrasts (usually Sundays) is identified by 0. The other groups are identified by 1, 2,... n (<= 6). For instance, usual trading days are defined by c(1,2,3,4,5,6,0), week days by c(1,1,1,1,1,0,0), week days, Saturdays, Sundays by c(1,1,1,1,1,2,0) etc. |
contrasts |
If true, the variables are defined by contrasts with the 0-group. Otherwise, raw number of days is provided. |
Aggregated values for monthly or quarterly are the numbers of days belonging to a given group. Contrasts are the differences between the number of days in a given group (1 to 6) and the number of days in the reference group (0).
Time series (object of class c("ts","mts","matrix")) corresponding to each group, starting with the 0-group (contrasts = FALSE)
or the 1-group (contrasts = TRUE).
More information on calendar correction in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/a-calendar-correction
# Monthly regressors for Trading Days: each type of day is different # contrasts to Sundays (6 series) regs_td <- td(12, c(2020, 1), 60, groups = c(1, 2, 3, 4, 5, 6, 0), contrasts = TRUE) # Quarterly regressors for Working Days: week days are similar # contrasts to week-end days (1 series) regs_wd <- td(4, c(2020, 1), 60, groups = c(1, 1, 1, 1, 1, 0, 0), contrasts = TRUE)# Monthly regressors for Trading Days: each type of day is different # contrasts to Sundays (6 series) regs_td <- td(12, c(2020, 1), 60, groups = c(1, 2, 3, 4, 5, 6, 0), contrasts = TRUE) # Quarterly regressors for Working Days: week days are similar # contrasts to week-end days (1 series) regs_wd <- td(4, c(2020, 1), 60, groups = c(1, 1, 1, 1, 1, 0, 0), contrasts = TRUE)
Canova-Hansen test for stable trading days
td_canovahansen( s, differencing, kernel = c("Bartlett", "Square", "Welch", "Tukey", "Hamming", "Parzen"), order = NA )td_canovahansen( s, differencing, kernel = c("Bartlett", "Square", "Welch", "Tukey", "Hamming", "Parzen"), order = NA )
s |
a |
differencing |
Differencing lags. |
kernel |
Kernel used to compute the robust covariance matrix. |
order |
The truncation parameter used to compute the robust covariance matrix. |
list with the ftest on td, the joint test and the details for the stability of the different days (starting with Mondays).
s <- log(ABS$X0.2.20.10.M) td_canovahansen(s, c(1, 12))s <- log(ABS$X0.2.20.10.M) td_canovahansen(s, c(1, 12))
Residual Trading Days Test
td_f( s, model = c("D1", "DY", "DYD1", "WN", "AIRLINE", "R011", "R100"), nyears = 0 )td_f( s, model = c("D1", "DY", "DYD1", "WN", "AIRLINE", "R011", "R100"), nyears = 0 )
s |
a |
model |
the model to use for the residuals. See details. |
nyears |
|
The function performs a residual seasonality test that is a joint F-Test on the coefficients of trading days regressors. Several specifications can be used on the model:
model = "WN" the following model is used:
model = "D1" (the default) the following model is used:
model = "DY" the following model is used:
model = "DYD1" the following model is used:
model = "AIRLINE" the following model is used:
model = "R011" the following model is used:
model = "R100" the following model is used:
td_f(ABS$X0.2.09.10.M)td_f(ABS$X0.2.09.10.M)
Likelihood ratio test on time varying trading days
td_timevarying(s, groups = c(1, 2, 3, 4, 5, 6, 0), contrasts = FALSE)td_timevarying(s, groups = c(1, 2, 3, 4, 5, 6, 0), contrasts = FALSE)
s |
The tested time series |
groups |
The groups of days used to generate the regression variables. |
contrasts |
The covariance matrix of the multivariate random walk model used for the time-varying coefficients are related to the contrasts if TRUE, on the actual number of days (all the days are driven by the same variance) if FALSE. |
A Chi2 test
s <- log(ABS$X0.2.20.10.M) td_timevarying(s)s <- log(ABS$X0.2.20.10.M) td_timevarying(s)
Creates a time series object
to_ts(source, id, type = "All")to_ts(source, id, type = "All")
source |
Source of the time series |
id |
Identifier of the time series (source-dependent) |
type |
Type of the requested information (Data, Metadata...). All by default. |
An object of type "JD3_TS". List containing the identifiers, the data and the metadata
Creates a collection of time series
to_tscollection(source, id, type = "All")to_tscollection(source, id, type = "All")
source |
Source of the collection of time series |
id |
Identifier of the collection of time series (source-dependent) |
type |
Type of the requested information (Data, Metadata...). All by default. |
An object of type "JD3_TSCOLLECTION". List containing the identifiers, the metadata and all the series.
Computes trigonometric variables at different frequencies.
trigonometric_variables(frequency, start, length, s, seasonal_frequency = NULL)trigonometric_variables(frequency, start, length, s, seasonal_frequency = NULL)
frequency |
Frequency of the series, number of periods per year (12,4,3,2..) |
start, length
|
First date (array with the first year and the first period)
(for instance |
s |
time series used to get the dates for the trading days variables. If supplied the
parameters |
seasonal_frequency |
the seasonal frequencies. By default the fundamental seasonal frequency and all the harmonics are used. |
Denote by the value of frequency (= the period) and
, ..., the frequencies provides by seasonal_frequency
(if seasonal_frequency = NULL then and =i).
trigonometric_variables returns a matrix of size .
For each date associated to the period (),
the columns and are equal to:
Take for example the case when the first date (date) is a January, frequency = 12
(monthly time series), length = 12 and seasonal_frequency = NULL.
The first frequency, represents the fundamental seasonal frequency and the
other frequencies (, ..., )
are the five harmonics. The output matrix will be equal to:
Multiplicative adjustment of a time series for leap year / length of periods
ts_adjust(s, method = c("LeapYear", "LengthOfPeriod"), reverse = FALSE)ts_adjust(s, method = c("LeapYear", "LengthOfPeriod"), reverse = FALSE)
s |
The original time series |
method |
|
reverse |
Adjustment or reverse operation |
The interpolated series
y <- ABS$X0.2.09.10.M ts_adjust(y) # with reverse we can find the all.equal(ts_adjust(ts_adjust(y), reverse = TRUE), y)y <- ABS$X0.2.09.10.M ts_adjust(y) # with reverse we can find the all.equal(ts_adjust(ts_adjust(y), reverse = TRUE), y)
Interpolation of a time series with missing values
ts_interpolate(s, method = c("airline", "average"))ts_interpolate(s, method = c("airline", "average"))
s |
The original time series |
method |
airline: interpolation through an estimated airline model average: interpolation using the average of the previous and next non missing values |
The interpolated series
Title
tsdata_of(values, dates)tsdata_of(values, dates)
values |
Values of the time series |
dates |
Dates of the values (could be any date inside the considered period) |
A ts object. The frequency will be identified automatically and missing values will be added in need be.
The identified frequency will be the lowest frequency that match the figures.
The provided data can contain missing values (NA)
# Annual series s <- tsdata_of(c(1, 2, 3, 4), c("1990-01-01", "1995-01-01", "1996-01-01", "2000-11-01")) # Quarterly series t <- tsdata_of(c(1, 2, 3, NA, 4), c("1990-01-01", "1995-01-01", "1996-01-01", "2000-08-01", "2000-11-01"))# Annual series s <- tsdata_of(c(1, 2, 3, 4), c("1990-01-01", "1995-01-01", "1996-01-01", "2000-11-01")) # Quarterly series t <- tsdata_of(c(1, 2, 3, NA, 4), c("1990-01-01", "1995-01-01", "1996-01-01", "2000-08-01", "2000-11-01"))
Makes a UCARIMA model canonical; more specifically, put all the noise of the components in one dedicated component
ucarima_canonical(ucm, cmp = 0, adjust = TRUE)ucarima_canonical(ucm, cmp = 0, adjust = TRUE)
ucm |
An UCARIMA model returned by |
cmp |
Index of the component that will contain the noises; 0 if a new component with all the noises will be added to the model |
adjust |
If TRUE, some noise could be added to the model to ensure that all the components has positive (pseudo-)spectrum |
A new UCARIMA model
mod1 <- arima_model("trend", delta = c(1, -2, 1)) mod2 <- arima_model("noise", var = 1600) hp <- ucarima_model(components = list(mod1, mod2)) hpc <- ucarima_canonical(hp, cmp = 2)mod1 <- arima_model("trend", delta = c(1, -2, 1)) mod2 <- arima_model("noise", var = 1600) hp <- ucarima_model(components = list(mod1, mod2)) hpc <- ucarima_canonical(hp, cmp = 2)
Estimate UCARIMA Model
ucarima_estimate(x, ucm, stdev = TRUE)ucarima_estimate(x, ucm, stdev = TRUE)
x |
Univariate time series |
ucm |
An UCARIMA model returned by |
stdev |
TRUE if standard deviation of the components are computed |
A matrix containing the different components and their standard deviations if stdev is TRUE.
mod1 <- arima_model("trend", delta = c(1, -2, 1)) mod2 <- arima_model("noise", var = 16) hp <- ucarima_model(components = list(mod1, mod2)) s <- log(aggregate(retail$AutomobileDealers)) all <- ucarima_estimate(s, hp, stdev = TRUE) plot(s, type = "l") t <- ts(all[, 1], frequency = frequency(s), start = start(s)) lines(t, col = "blue")mod1 <- arima_model("trend", delta = c(1, -2, 1)) mod2 <- arima_model("noise", var = 16) hp <- ucarima_model(components = list(mod1, mod2)) s <- log(aggregate(retail$AutomobileDealers)) all <- ucarima_estimate(s, hp, stdev = TRUE) plot(s, type = "l") t <- ts(all[, 1], frequency = frequency(s), start = start(s)) lines(t, col = "blue")
Creates an UCARIMA model, which is composed of ARIMA models with independent innovations.
ucarima_model(model = NULL, components, complements = NULL, checkmodel = FALSE)ucarima_model(model = NULL, components, complements = NULL, checkmodel = FALSE)
model |
The reduced model. Usually not provided. |
components |
The ARIMA models representing the components |
complements |
Complements of (some) components. Usually not provided |
checkmodel |
When the model is provided and checkmodel is TRUE, we check that it indeed corresponds to the reduced form of the components; similar controls are applied on complements. Currently not implemented |
A list with the reduced model, the components and their complements
mod1 <- arima_model("trend", delta = c(1, -2, 1)) mod2 <- arima_model("noise", var = 1600) hp <- ucarima_model(components = list(mod1, mod2)) print(hp$model)mod1 <- arima_model("trend", delta = c(1, -2, 1)) mod2 <- arima_model("noise", var = 1600) hp <- ucarima_model(components = list(mod1, mod2)) print(hp$model)
Wiener Kolmogorov Estimators
ucarima_wk(ucm, cmp, signal = TRUE, nspectrum = 601, nwk = 300)ucarima_wk(ucm, cmp, signal = TRUE, nspectrum = 601, nwk = 300)
ucm |
An UCARIMA model returned by |
cmp |
Index of the component for which we want to compute the filter |
signal |
TRUE for the signal (component), FALSE for the noise (complement) |
nspectrum |
Number of points used to compute the (pseudo-) spectrum of the estimator |
nwk |
Number of weights of the Wiener-Kolmogorov filter returned in the result |
A list with the (pseudo-)spectrum, the weights of the filter and the squared-gain function (with the same number of points as the spectrum)
mod1 <- arima_model("trend", delta = c(1, -2, 1)) mod2 <- arima_model("noise", var = 1600) hp <- ucarima_model(components = list(mod1, mod2)) wk1 <- ucarima_wk(hp, 1, nwk = 50) wk2 <- ucarima_wk(hp, 2) plot(wk1$filter, type = "h")mod1 <- arima_model("trend", delta = c(1, -2, 1)) mod2 <- arima_model("noise", var = 1600) hp <- ucarima_model(components = list(mod1, mod2)) wk1 <- ucarima_wk(hp, 1, nwk = 50) wk2 <- ucarima_wk(hp, 2) plot(wk1$filter, type = "h")
Allows to combine two or more calendars into one calendar, weighting all the holidays of each of them.
weighted_calendar(calendars, weights)weighted_calendar(calendars, weights)
calendars |
list of calendars. |
weights |
vector of weights associated to each calendar. |
Composite calendars are useful for a series that including data from more than one country/region. They can be used, for example, to create the calendar for the European Union or to create the national calendar for a country, in which regional holidays are celebrated. For example, in Germany public holidays are determined by the federal states. Therefore, Epiphany is celebrated only in Baden-Wurttemberg, Bavaria and in Saxony-Anhalt, while from 1994 Day of Repentance and Prayer is celebrated only in Saxony.
returns an object of class c("JD3_WEIGHTEDCALENDAR","JD3_CALENDARDEFINITION")
More information on calendar correction in JDemetra+ online documentation: https://jdemetra-new-documentation.netlify.app/a-calendar-correction
national_calendar, chained_calendar
Belgium <- national_calendar(list(special_day("NEWYEAR"), fixed_day(7, 21))) France <- national_calendar(list(special_day("NEWYEAR"), fixed_day(7, 14))) composite_calendar <- weighted_calendar(list(France, Belgium), weights = c(1, 2))Belgium <- national_calendar(list(special_day("NEWYEAR"), fixed_day(7, 21))) France <- national_calendar(list(special_day("NEWYEAR"), fixed_day(7, 14))) composite_calendar <- weighted_calendar(list(France, Belgium), weights = c(1, 2))