Hello,
I want to run IRT and Mantel-Haensel procedures as a parametric bootstrap by resampling (with replacement) from my dataset with the probability of selecting each observation proportional to the survey weights. I was told that this may allow me to carry out the procedures robust to survey weighting and clustering of survey weights due to complex sample design.
My questions for the Statalist community:
- How do I specify probability of selection proportional to the survey weights?
- What would the code look like for running this in Stata? Do you recommend using the bsample or bootstrap estimation procedures? Ideally, we would like to do 1000 reps, so, I believe the bootstrap estimation would be appropriate for this?
- Since Stata does not allow svy weights with Mantel-Haensel, is this a feasible endeavor?
Thank you!
Helena
I want to run IRT and Mantel-Haensel procedures as a parametric bootstrap by resampling (with replacement) from my dataset with the probability of selecting each observation proportional to the survey weights. I was told that this may allow me to carry out the procedures robust to survey weighting and clustering of survey weights due to complex sample design.
My questions for the Statalist community:
- How do I specify probability of selection proportional to the survey weights?
- What would the code look like for running this in Stata? Do you recommend using the bsample or bootstrap estimation procedures? Ideally, we would like to do 1000 reps, so, I believe the bootstrap estimation would be appropriate for this?
- Since Stata does not allow svy weights with Mantel-Haensel, is this a feasible endeavor?
Thank you!
Helena
