topic Fixed Effect and biased estimates in Statistical Procedures

Problem in fixed effect estimates

Saba1 — Fri, 29 Sep 2017 22:26:43 GMT

I have a panel dataset with a large number of stocks and monthly data. I have tried to run regression with "year" and "stock" fixed effect by using proc glm. But the results show biased estimates for year and stock. Following is the message I receive with regression output:

"Note: The X'X matrix has been found to be singular, and a generalized inverse was used to solve the normal equations. Terms whose estimates are followed by the letter 'B' are not uniquely estimable."

Following is the code:

proc glm data=reg ;
absorb year PERMNO;
model a = b c year PERMNO / solution noint;
run;
quit;

Please assist me identifying the reason behind biased results. Is it fine if I use "proc surveyreg" instead of proc glm for fixed effect?

Thanks.

Fixed Effect and biased estimates

Saba1 — Fri, 29 Sep 2017 22:31:48 GMT

Following is the code:

proc glm data=reg ;
absorb year PERMNO;
model a = b c year PERMNO / solution noint;
run;
quit;

Please assist me identifying the reason behind biased results. Is it fine if I use "proc surveyreg" instead of proc glm for fixed effect?

Thanks.

Re: Fixed Effect and biased estimates

Kurt_Bremser — Sat, 30 Sep 2017 06:23:53 GMT

I merged the questions. Please post questions omly once, as it keeps all the answers and discussion in one place.

The stat procs forum is the right one for this issue.

Re: Fixed Effect and biased estimates

Saba1 — Thu, 05 Oct 2017 22:07:18 GMT

@Kurt_Bremser Thanks. But I have not yet received any help.