10-06-2016 11:42 AM
I have a firm-date panel and wanted to generate predict value (beta hat) for s_score at the date level. So far, I have the below and wanted to ask you if you knew how to do this.
proc sort data=b5; by date; run;
proc reg data=b5; by date; model mtb = size roa s_score/adjrsq;
output out=o1 p=p residual=r; run; quit;
when i do this, i get a p value next to every single firm date obs.
essentially, i want a b to s_score at the date level.
Thank you in advance,
10-06-2016 12:33 PM
You should also have parameter estimate for each date as well in the Parameter estimates.
Do you mean that you want the parameter in an output data set? That would be from using the OUTEST= option on the Proc Reg statement.
10-07-2016 09:30 AM
To expand on @ballardw's reply, the statistics that you are looking for are called "parameter estimates." That should help you with internet searches.
In addition to using the OUTEST= option, you should add the NOPRINT option to the PROC REG statement so that the screen is not filled with hundreds of tables/graphs that you don't need to see. The following statements create an output data set called ParamEst that contains all parameter estimates. You can use a WHERE clause to restrict the output it you only want the estimates for s_score.
proc reg data=b5 outest=ParamEst noprint; by date; model mtb = size roa s_score / adjrsq; quit;