Robust Standard Errors when using Mundlak approach

[Kevin van Dam]

Hi all,

For my panal data analysis i'm using a mundlak model. I'm using the commands Mundlak with the option Hybrid. To correct for serial autocorrelation (according to a significant xtserial test), i should use robust standard errors. However, the option vce(robust) is not possible when using the mundlak command. Is there a way to still use robust standard errors? Or does the mundlak command automatically corrects for serial autocorrelation? Thanks!

[JJ Kovach]

I believe the xthybrid command can also perform the Mundlak transformation and utilize robust standard errors

[Kevin van Dam]

Thaks for your answer. However, It takes approx 5 minutes to load the regression results if I run the xthybrid command. That's why i'm using the mundlak with option hybrid command.

[Jeff Wooldridge]

It's easy to implement the Mundlak regression directly after obtaining the time averages -- especially for a balanced panel. You can then use -reg- with the vce(cluster id) option. With an unbalanced panel, you need to use only the time periods with complete cases for each cross-sectional unit when computing the time averages.