Hi Zahid, 

If I remember correctly you can estimate a Semipar in R.  There is a package in R called semipar . here it is https://cran.r-project.org/web/packages/SemiPar/SemiPar.pdf 

What do you want to compare ?

Dear Yashar

Thank you very much. I want to compare impact of income on calorie demand. Since this relationship may be non-linear or semi-parametric as well. So I want to compare linear, semi-parametric, semi-parametric Single index and non-parametric model performance for calorie response to income (some other regressors like HHsize, urban/rural etc) are also in the model. I dont know how to compare Semipar Single index with other models.

Hi Zahid, you can look up the 'np' library written by Jeff Racine. This provides a general non-parametric test for correct model fit and also allows estimation of fully non-parametric models, partially linear models and index models. (you might also need the 'crs' package). You could then do all your estimations in R followed by a robust test for model fit. I'm pretty sure it will do what you are after but not absolutely - just check the vignette i guess. 

Hope that helps.

-Daniel

Thanks Daniel, Racine book is useful for our analysis. But may be we are not good in R so finding it difficult to program exactly. We compared all others using approach you are suggesting but main issue is with Semi-parametric Single Index model.

Regards

Zahid