Random Effect Autocorrelation Test

[David Junius]

Dear Good People,

I am currently shifting to using Random Effect for my data and I would like to know if there is a spesific command to execute autocorrelation test using Random Effect? Previously I use the command "xtserial y1 x1 x2". And also, I use robust for my Random Effect, is there any different in the command? Thank you!

[Carlo Lazzaro]

David:
you can still use the community-contributed command -xtserial-.

[David Junius]

Thank you again Carlo Lazzaro !

[David Junius]

Dear Carlo Lazzaro , I have a few questions: I've tried using robust on my RE model, but the result of autocorrelation and hetero tests show that the problem still occur, is it absolute that using robust will treat autocorrelation and hetero that we don't need to run the test again? can you share any journal regardings this as I also need it for my bibliography.

Second Question, I am suggested to use xtpsce but I'm still looking for information regardings the command, can you explain to me if xtpsce is exclusive to either FE or RE model or can be applied to both? and if xtpsce will treat hetero and autocorrelation just like robust?
Thankyou so much for your help

[Carlo Lazzaro]

David:
1) if you impose cluster-robust SE, there's no scope in repeating the tests for autocorrelation and/or heteroskedasticity: those nuisances still exist, but the regression model will deal with them;
2) -xtpcse- is a pooled OLS estimator for long T, short N panel datasets. It treats heteroskedastciity and autocorrelation togehter or separately. Methods and formulas of -xtpcse- related entries can be helpful for comparison with -xtreg-.

[David Junius]

Dear Carlo Lazzaro ,

Do you have any research studies or journal that you can share to me regardings how robust can treat hetero and autocorrelation? I need them for citation and bibliography (and as a prove to my professor). Thanks again!

[Carlo Lazzaro]

David:
among other textbooks, take a look at https://www.stata.com/bookstore/micr...metrics-stata/

[David Junius]

Thank you so much Carlo Lazzaro

[Matt Joyce]

Hi Carlo Lazzaro . I noticed you suggested using xtserial to test for autocorrelation in fixed and random effects models. After a bit of research, I have found that this command cannot be used for time variant models. I must present some results to show whether my model has autocorrelation. Do you know if there is a test for autocorrelation that can be applied to two-way models (for both fixed and random effects)?

[Carlo Lazzaro]

Matt:
see https://www.statalist.org/forums/for...al-correlation and https://journals.sagepub.com/doi/pdf...867X0300300206