Hi there,

I was wondering if someone could clear up a problem I have. I have to review the assumptions of a fixed effects model of a certain study.

One of the assumptions we learned for efficiency of a FE model, is that the error terms in different time periods are uncorrelated (called serial correlation). In the article they don't talk about this, and we don't have the data to check for ourselves.

Let's say there is serial correlation. How do you overcome this? I read in other discussions here that this can be overcome by clustering at the identifying unit variable.

They use a stratified sample, with the strata being on the school, and the unit being individuals. They use standard errors that are clustered by the sampling stratum. I believe they mean that the SE are clustered at the school identifying variable. Or am I interpreting this wrong?

Would this overcome serial correlation as well? Since the clustering is not happening at the unit variable (assuming my interpretation is correct).


Thank you in advance (: