Applying longitudinal weights to -xt- commands (panel data)

[Claire James]

Hi Statalist,

I'm currently facing an issue whereby I'm not able to apply longitudinal weights to my -xt- commands using the -svy- command, e.g. having specified the survey design beforehand, I can't run:

svyset ER31997 [pweight=weights], strata(ER31996)
svy: xtlogit emp EBreceived UEduration EBduration stateUErate i.month, fe nolog

Because -svy- commands do not support -xt- commands. Is there a way to apply weights directly to my -xt- regressions to bypass this issue? An issue that I foresee with this is that -xt- regressions require weights to be constant within the panel, but my weights are time-varying. From searching up this problem online before this, it seems that a potential solution is using the GLLAMM programme, but I don't think I will be able to use it as it is apparently computationally demanding and slow.

Any suggestions would be much appreciated!

[Clyde Schechter]

Unless you are using a very old version of Stata, the -melogit- or -meglm- command should do what you were thinking of doing using GLLAMM, and these commands are supported by the -svy:- prefix. You won't get a fixed effects model, you'll get random effects. But that's true of GLLAMM, too.

[Claire James]

Thanks Clyde!