Hi all,
I am working on a panel dataset of birth histories 1988-2005. I want to keep only those women (woman is my unit of obs) that didn't change residence since 1988, the Stayers. However, the Stayers are different on multiple characteristics such as age, wealth, education from the Movers, that are the rest of the sample.
To use the subsample of stayers and make it representative of the average individual in the sull sample, I am exploiting propensity score matching. I compute weights based on propensity score and then I use them when running regressions in my subsample of stayers.
As you can seen below, I am using inverse probability weights:
local characteristics "age_woman WealthIndex Education children urban number_of_males total_abortions v701"
xtset woman year
logit stayers88 `characteristics'
predict p_stayer, xb
gen weightPSM = 1 / p_stayer
My question is: is this approach correct even if I am using panel data? is logit correct or should I use xtlogit?
Many thanks!
I am working on a panel dataset of birth histories 1988-2005. I want to keep only those women (woman is my unit of obs) that didn't change residence since 1988, the Stayers. However, the Stayers are different on multiple characteristics such as age, wealth, education from the Movers, that are the rest of the sample.
To use the subsample of stayers and make it representative of the average individual in the sull sample, I am exploiting propensity score matching. I compute weights based on propensity score and then I use them when running regressions in my subsample of stayers.
As you can seen below, I am using inverse probability weights:
local characteristics "age_woman WealthIndex Education children urban number_of_males total_abortions v701"
xtset woman year
logit stayers88 `characteristics'
predict p_stayer, xb
gen weightPSM = 1 / p_stayer
My question is: is this approach correct even if I am using panel data? is logit correct or should I use xtlogit?
Many thanks!
