nbbrd / jdemetra-r Goto Github PK

Hello.

I have a question :

I succeed in generating new specifications which overwrite some of the options of a pre-specified specification.

For example, I have no problem to constraint a log-transformation on data :

spec_multi = spec_create() spec_str(spec_multi,"regarima.transform.function","Log")

But to modify options linked to dates, I have some difficulties. More precisely: I have a time serie, which start in 01-01-1998 and end to 01-01-2017, and I would like to limit the estimate model span from 01-01-2009 to 01-01-2017. I haved tested a lot of way, without any success.

I will be grateful for any help,

Kévin

(Excellent plug-in by the way !)

Could someone help me to understand what the result in jd-testsa.R mean? I tried to understand the result by compare it with the result in the Jdemetra app, but I can't get it
Hopefully someone can help me please

When using the jd_td function I have tried to include two regressors using a formula such as y~x+z. I think it's only using the first variable. I have looked at the code and found the following line which seems to take the first series in the formula

xvar<-jd_tsvar(get(X.series.names[1], envir=environment(X.formula)), X.series.names[1])

Is it possible to include more than one regressor?

Hi,

When trying to insert an additive outlier or a large extreme at a particular date, I tried the following code
spec_strs(spec, "regarima.regression.outliers", c("AO.2005-12-01.f"))
but it didn't do anything. What is the correct format for this? Also, can I set an outlier to be a particular value?

Likewise, when trying to adjust the window length for moving trading day, I tried the following code
spec_int(spec1,"regarima.regression.tradingdays.WindowLength",15)
but it didn't do anything. What is the correct format for this? Also, how do I set the smoother type and whether or not to re-estimate the ARIMA model?

Thanks,
Peter.

I have developed a function to remove user defined regressors - similar to jd_unregistercalendar. It may be useful to include in the package.

jd_unregisterVariable<- function(name, group){
jd_context<-.jcall("ec/tstoolkit/algorithm/ProcessingContext", "Lec/tstoolkit/algorithm/ProcessingContext;", "getActiveContext")
jd_vars<-.jcall(jd_context, "Lec/tstoolkit/timeseries/regression/TsVariables;", "getTsVariables", group)
.jcall(jd_vars, "Z", "remove", name)
}

I am trying to run the jd_test.R file but I get an error message suggesting that some necessary files do not exist, which is clearly not true.

I have tried specifying the exact address but it also doesnt work
source("C:/Users/David/Documents/GitHub/jdemetra-R/jd_init.R")

Hi,

I'm unsure whether this is an issue for here or for the Java code - I suspect the issue may lie in the Java code, but I found it when running this code, so I'll report it here.

My attached code (apologies for the extra .txt at the end, GitHub won't take an R file) creates four specifications which are all identical except two have Easter regressors and two don't, and two have moving trading day and two don't. For the two without moving trading day, one has the Easter regressors and the other one doesn't, so the seasonally adjusted series from each should be different, which they are. For the two with moving trading day however, one has the Easter regressors and the other one doesn't, so the seasonally adjusted series from each should be different, but they're not. So it appears that turning on moving trading day is somehow disabling user defined regressors.
Testing JD+ R regression functionality.R.txt

I've used a file with regressors for Easter (with linearly increasing effect for the 7 days prior to Easter and constant effect for the 4 days of the Easter long weekend, as two separate regressors) as I didn't want to use a regressor that was associated with trading day. I put it in the data folder for running the code.
EasterQuL_74.txt

Thanks,
Peter.