topic Re: Stationarity in SAS Forecasting and Econometrics

Stationarity

Lok07 — Fri, 12 Apr 2019 15:14:37 GMT

Dear community

I have a monthly time serire data and i want test the stationarity of this serie.

Can one give me any suggestion!!

regards

Re: Stationarity

imvash — Fri, 12 Apr 2019 15:29:00 GMT

If you use SAS Viya, you can use TSA.STATIONARITYTEST function. See the example below (I copied from documentation), which uses sashelp.air data set in CAS:

proc tsmodel data=mycas.air outscalar=mycas.outscalars;
   id date interval=month;
   var air;
   outscalars stationary1 stationary2;
   require tsa;
   submit;
   declare object TSA(tsa);
   stationary1=1; stationary2=1;
   rc = TSA.STATIONARITYTEST(air,,,,,pvalue);
   *test with the default significant level=0.05;
   if rc =1 then stationary1 = 0;
   *test with significant level = 0.1;
   if pvalue > 0.1 then stationary2 = 0;
   endsubmit;
run;

You can get more information regarding the TSA package and this function at

https://go.documentation.sas.com/?docsetId=fstsp&docsetTarget=fstsp_tsa_sect091.htm&docsetVersion=15.1&locale=en

If you do not use SAS Viya, take a look at http://support.sas.com/documentation/cdl/en/etsug/63939/HTML/default/viewer.htm#etsug_macros_sect007.htm

Hope it helps!

Re: Stationarity

Lok07 — Fri, 12 Apr 2019 15:48:17 GMT

I don't use SAS viya.

I used the ADF and KPSS test.

Based on their output the time serie has a trend.

But when i want remove the trend using proc reg

Proc reg data=database;

model y= T;

run;

T is the date variable from Jan 2015 to Dec 2018.

data database;

set database;

TY= y-B*T;

run;

The TY which i created has no value.

I don't know how i can remove the trend of my variable.

Re: Stationarity

noetsi — Fri, 12 Apr 2019 21:56:56 GMT

There are two types of non-Stationarity. The first is stochastic non-Stationarity (which I think is what one uses the ADF and KPSS test for). This is dealt with by differencing.

The second type is a deterministic non-Stationarity. This is commonly dealt with by specifying a quadratic or cubic time variable.

You have to know which type of non-Stationarity you have and specify the right correction.

Re: Stationarity

Lok07 — Sun, 14 Apr 2019 22:26:28 GMT

Hi noetsi

I have a deterministic non-Stationarity.

Should I regress my variable on the date (date is from Jan 2015 to Dec 2018) variable like that

proc reg data=data;

model Y=date;

run;