Quantile regression (multilevel data)

[Nerea Gomez]

Good morning.

I wanted to ask if anyone can help me and give me information about which commands I should use for the following.

I am working with PISA data, which has a multi-level structure, as schools are selected and then students within the school participate in the test. In my specific case I am estimating the impact on academic performance of a number of variables, controlling for variables at the student level and also at the school level. I have made estimates with multilevel models (meglm) and now I want to see if the impact is different according to the performance tertile of the student. Therefore, I want to estimate "quantile regressions" but I have several doubts:

1. Is it possible to make quantile regressions that take into account this multilevel structure?

2. Is it possible to include the weights at student and school level in the quantile regressions?

Thank you in advance.

[Yuting Li]

Hello Nerea, are you still working on this and look for suitable model? I also have similar questions as you. Luckily, I found STATA 18.5 has updated a new command for Bayesian asymmetric Laplace model, which uses Bayesian estimation for quantile regression and can be extended to multi-level data (both random intercept and slope). See: https://www.stata.com/statanow/bayes...laplace-model/. Perhaps it will be helpful for you.