Forecasting using SAS Forecast Server, SAS/ETS, and more

PROC AUTOREG - SLSTAY Not Working

Reply
Highlighted
Frequent Learner
Posts: 1

PROC AUTOREG - SLSTAY Not Working

 

I'm estimating a time series autoregressive error model with PROC AUTOREG, using the slstay and backstep option to define the significance level of AR terms to keep in the model. However, slstay does not work as I understand it. For example, if I set slstay = 0.1, on occasion the output will include AR terms with p-values > 0.1. Change the value of slstay has no effect, until slstay = 0.06, when it removes those AR terms with p-values > 0.1.

 

Since this is utilizing a k-fold estimation process, I cannot change the significant level of slstay for each iteration or model I put through the simulation.

 

	proc autoreg data=to_forecast plots=all outest=parms_out_;
		where prodcat = "&prodcat.";
		model DV =  &model_vars.
		 lag_index lag_DV

		/ &noint_var.  method = ml nlag = 24 backstep maxiter=150 slstay=0.1 lagdep=lag_dif_rate;
		&restrict_st.

		/*OLS Estimates: */
		ods output Autoreg.Model1.OLSEst.ParameterEstimates = stderrs_probts_;
		ods output Autoreg.Model1.OLSEst.FitSummary = fitsum;

		/*ML Estimates (If AR terms): */
		ods output Autoreg.Model1.FinalModel.ParameterEstimatesGivenAR = stderrs_probts_ml;
		ods output Autoreg.Model1.FinalModel.FitSummary = fitsum_ml;
		
		ods output "Miscellaneous Statistics" = durbinh;
	run;

 

Any advice would be appreciated!

 

(I'm using EG 7.1 and Base SAS 9.4).

Ask a Question
Discussion stats
  • 0 replies
  • 227 views
  • 0 likes
  • 1 in conversation