topic Re: macro for a rolling window regression in SAS Procedures

macro for a rolling window regression

sasphd — Wed, 19 Nov 2014 13:24:50 GMT

I am working on a research project and have trouble come up with a way to conduct a rolling window regression. In my proc model I include this macro to do my rolling window.

%do fen = %eval(&dfen) %to 100;

where %eval(&fen) le obs le %eval(36+&fen-1);

However, my proc model is done by id and as you can see in this macro I suppose that the number of observations is 100 for each id but this is not true I have a different number of observation for each id. In my database I add a column which indicates the number of total number of observation for each id (nb).

How can I change the fixed number 100 to a changing variables called nb?

I try to do it like this :

%do fen = %eval(&dfen) %to %eval(&nb);

where %eval(&fen) le obs le %eval(36+&fen-1);

but I have this error

ARNING: Apparent symbolic reference NB not resolved.

ERROR: The text expression &NB contains a recursive reference to the macro variable NB. The

macro variable will be assigned the null value.

Re: macro for a rolling window regression

ballardw — Wed, 19 Nov 2014 15:56:37 GMT

This forum works best when only a single thread is opened for a give topic. If a suggestion is made to post in another subject area that's okay, but generally multiple threads are not needed.

Re: macro for a rolling window regression

art297 — Wed, 19 Nov 2014 21:51:16 GMT

In case my example wasn't sufficiently clear, here is the same type of code for doing proc reg on sashelp.class using rolling windows of 5 records within each sex. BTW, you appear to never mark your questions as having been answered, thus many of us continue to spend probably unnecessary time on them:

filename runit temp;

data _null_;

length script $255;

file runit;

set test (keep=sex);

by sex;

if first.sex then do;

grp=0;

counter=1;

end;

else counter+1;

if counter ge 5 then do;

grp+1;

script=catt('Title "Predicting age by height and weight -- Sex=',

sex,' group=',grp,'";');

put script;

script=catt("proc reg data=test (firstobs=",_n_-4," obs=",_n_,

"); model age=height weight; run;");

put script;

end;

run;

%include runit;

Re: macro for a rolling window regression

sasphd — Wed, 19 Nov 2014 22:55:23 GMT

thanks for your help.