How to apply longitudinal weights? Edited

Dear all,

I am using data from a rotational panel following households. Thus, households are followed for a number of years, but part of them is replaced by new families each year.

I deal with this structure as an unbalanced panel, but I absolutely need to use longitudinal weights to account for attrition and replacement. Weights are indeed time-varying, with the following structure.


ID | Year | Weight | Birth | Other covariates

1 2001 0.5 1 ............
1 2002 0.6 0
1 2003 0.7 0
2 2001 0.4 0
2 2002 0.3 0
2 2003 0.8 0
3 2002 0.7 0
3 2003 0.9 1
3 2004 0.3 0

My dependent variable is binary and represents the event of a birth within the household (1= yes, 0=no).


1) I tried to use a random effect panel regression through xtlogit through iweight

xtlogit Birth covariates [iweight=1/Weight], re

but this command does not allow me to use time-varying weights.

2) I also tried to use gllamm

gllamm Birth covariates, weight(Weight)

but it does not run the regression as well (I think it is because of the time-varying issue because gllamm runs the regression perfectly if weights do not vary over time).

Has someone any idea on how to overcome this issue?


Thank you very much,
Lydia