|
Hello again,
Another question. Is it possible to test the differences in parameter
estimates if the estimates come from the same model?
Can I just merge the estimates, then calculate the differences, and
then test whether the mean is zero. For some reason I don't think a
mean test of the differences is the way to go.??????? If anyone could
confirm this or inform me on this it would be great.
For example I have:
proc reg data=fileA outest = in.file1995 (keep=id year);
where year=1995;
model y = x1 x2;
run;
proc reg data=fileA outest = in.file1996 (keep=id year);
where year=1996;
model y = x1 x2;
run;
data in.merge_master;
Merge in.file1995 in.file1996;
by id year;
run;
data in.merge_master;
set in.merge_master;
diff = est1995 - est1996
run;
proc mean data = in.merge_master mean stddev median;
var diff;
run;
Any suggestions would be greatly appreciated.
| 
