Testing for autocorrelation with a panel dataset

[Katerina Novakova]

Dear Stata Forum,

I am trying to figure out how to test for autocorrelation of my panel dataset (of 2 periods), using the RE model in my regression.

I Googled, ChatGPTd, and also read some of the discussions here, which mostly tried to suggest the -xtserial- command; however, this command does not work for my T=2. Some other sources suggest ways how to do it with FE and suggest that -vce(robust)- already helps with autocorrelation (apart from ensuring homoskedasticity) in the FE model. But would that also be the case for an RE model? Are there any other ways that I should consider (keeping in mind my use of RE and T=2)?

Thank you once again for your help.

Katerina

[Andrew Musau]

You cannot test for serial correlation with \(T=2\) and you don't need to do it in \(N >> T\) panels. If RE is consistent, go ahead and estimate the model with -vce(cluster panelvar)-.

[Katerina Novakova]

Perfect,

Thank you for the clarification, Andrew.