run a regression with fixed effect and clustering, but year fixed effect in the result is strange

superclj123 — Mon, 13 Sep 2021 04:18:44 GMT

Dear all

I run a regression with both years fixed effect and cluster standard error by firms, the result is below. I found the year fixed effect is so strange, why there are only two years. My data period is from 1990 to 2019. Can someone tell me what's wrong? I wrote down the code here.

proc surveyreg data=tmp1.dataset; class year;
cluster permno;
model future_accuracy = rw3_std varx vary varx*vary L_abnret AbnRet firm_age size lev BTM rw10_std busseg sdret earn year / solution;
run;quit;

Also the degree of freedom here is strange, I have 20000 observations in my dataset but the degree of freedom here is only about 1000?

Anyone can help? Thanks in advance

Re: run a regression with fixed effect and clustering, but year fixed effect in the result is strang

ballardw — Mon, 13 Sep 2021 16:11:49 GMT

Post the LOG from running the code.

That will indicate how many records are used.

It may be that only those two years you've shown have non-missing values for ALL of the variables on your model statement. By default any missing value for a variable on a model statement will result result in the record being excluded from the analysis. Which likely impact the degrees of freedom as well.

topic run a regression with fixed effect and clustering, but year fixed effect in the result is strange in SAS Programming

run a regression with fixed effect and clustering, but year fixed effect in the result is strange

Re: run a regression with fixed effect and clustering, but year fixed effect in the result is strang