Difference between Quantile Regression and npregress

[Jessica Li]

Hi,
I am running into a problem about model selection. I am working on a validation of a questionnaire score: propr. I want to compare the propr with two other scores (sf6d and meanhs). My design is to fit the propr score as a function of binary exposure variables when accounting for age and sex as confounding variables. And do the same thing to the other scores.
Since the distribution of one score is skewed, I need to conduct a nonparametric regression. My questions are

1. Is quantile regression or npregress model a better for current scenario? The observation is around 300, and npregress was working on STATA.
2. If I choose quantile regression, is it possible to use median as the quantile for all the models?

Hope I stated my question clearly. Thanks!

[FernandoRios]

Hi Jessica,
Not completely sure what you want to do here. But let me rephrase it, what you want is to obtain the best linear or nonlinear predictor for propr, sf6d and meanhs. Correct?
Well, if that is the case, perhaps the best option is npregress.
Quantile regression is better suited to obtain more than one prediction for any combination of your explanatory variables. Using only the median would be almost like using a linear regression (and estimating the conditional mean), and it seems that is not what you want.
HTH
Fernando

[Jessica Li]

Thanks for your help, Fernando.
Yes, you understand my question correctly. I am going to use npregress!