Hi experts!

Can I use the STATA command lpoly on a binary outcome variable? The examples on lpoly's help file are all about continuous depvars. I do find some relevant literature online that say yes, but they are too technical and I cannot understand. Do I have to transform my binary depvar into a logit? Can I just use the raw depvar in a lpoly regression?

I know there is another new command npregress that can be safely applied to binary depvar (https://blog.stata.com/2017/06/27/no...ssion-but-not/). But in my research, I need to run local regression of yvar (binary) on a xvar (continuous) for two different samples, and then calculate the difference in the effect size of the xvar on the probability of (Y = 1) between the two samples AND also get the confidence interval of this difference, over the whole range of xvar. Insofar as I understand, npregress only gives us the confidence interval of the effect size of xvar on the local mean of yvar (which I assume is equal to the local probability as in my case the yvar is a binary var) within a given sample, while what I finally need is the CI of the difference of the effect sizes of xvar between two samples. Is there any way of obtaining this quantity using npregress?

Thank you very much!