Dear all,
I am investigating change in mental health between two measurement points and I would like to use inverse probability weighting to account for nonparticipation. So, I am running the analysis on those who participated at T1 and T2, which is about 50 percent of those who participated in T1. Among those who particiapted in both surveys I find that there are an underrepresantation of boys and older participants. I have therefore created the following pweight:
logit Participation i.Sex c.Age
predict P, p
gen IPW=1/P
I would like to use this weight in the following mixed command:
mixed DP i.Time##i.Sex c.Age || PID:, covariance(identity) mle stddeviations vce(robust)
As I understand, and as I have seen in previous posts, weights can be included at both levels as follows:
mixed DP i.Time##i.Sex c.Age [pweight=IPW] || PID:, pweight(IPW) covariance(identity) mle stddeviations vce(robust)
However, since I only have one weight, I am wondering if I should only adjust for nonparticipation at one level. So, my question is simply whether I should include this weight at both levels or if I should include it at one, and if so, which level?
This is the first time I am using weights so any answer is much appriciated.
Kind regards,
Bodil
I am investigating change in mental health between two measurement points and I would like to use inverse probability weighting to account for nonparticipation. So, I am running the analysis on those who participated at T1 and T2, which is about 50 percent of those who participated in T1. Among those who particiapted in both surveys I find that there are an underrepresantation of boys and older participants. I have therefore created the following pweight:
logit Participation i.Sex c.Age
predict P, p
gen IPW=1/P
I would like to use this weight in the following mixed command:
mixed DP i.Time##i.Sex c.Age || PID:, covariance(identity) mle stddeviations vce(robust)
As I understand, and as I have seen in previous posts, weights can be included at both levels as follows:
mixed DP i.Time##i.Sex c.Age [pweight=IPW] || PID:, pweight(IPW) covariance(identity) mle stddeviations vce(robust)
However, since I only have one weight, I am wondering if I should only adjust for nonparticipation at one level. So, my question is simply whether I should include this weight at both levels or if I should include it at one, and if so, which level?
This is the first time I am using weights so any answer is much appriciated.
Kind regards,
Bodil
