Adding Weights to Regression Using Panel Data

[Emma Kemp]

Hi there,

I am currently using the UKHLS Understanding Society panel dataset. In particular I have merged waves A, C, F I and L.
I am interested in the impact of a change in perceived social cohesion on life satisfaction.

- lfsato: life satisfaction
- NSC_index_nm: a balanced index consisting of 8 variables.

I am trying to incorporate sample weights to circumvent biased estimates.
UKHLS suggest using stata's svy suite ofcommandsso I ran the following:

[svyset, clear
svyset psu [pweight = l_indscus_lw], strata(strata)
svy: xtreg lfsato NSC_index_nm i.wave, fe robust ]

However, an error message appears that says xtreg is not supported by svy.

Do you know of any commands that will work to run regressions on panel data incorporating probability weights?

I have read that areg might work. Would the following code achieve what I am trying to do?

[areg lfsato NSC_index_nm i.wave [aweight=l_indscus_lw], absorb(pidp) cluster(psu) // Use areg for fixed effects regression with weights and clustered standard errors - errors clustered around the primary sampling unit (households)]

Thank you in advance for any advice you might be able to provide.

Emma

[Rich Goldstein]

I think you will need to do this via -sem- or -gsem-; you can start with examples 38g and 39g in

help sem_examples

[Emma Kemp]

Thank you very much Rich!