Hi everyone,
I am trying to do a nonparametric regression using -npregress- command, but Stata gave the same bandwidth when I apply Biweight, Triangle and Epan2 kernel function with the default cross validation bandwidth selection.
The following is my code and output
In my regression, ptexp is potential experience, it is an integer variable which can take value from 1 to 20. afqt is an standardized test score which is continuous.
I use the exactly the same data set for each nonparametric regression. The only difference is the choice of kernel function.
The big issue here is that the bandwidth for ptexp is 4.004203 and for afqt is 0.7226754 all the time no matter which kernel function I use.
I don't understand why I got the same bandwidths even if I chose different kernel functions. Is it possible that I may get the same bandwidth no matter which kernel function I used?
Thank you so much
Xizi
I am trying to do a nonparametric regression using -npregress- command, but Stata gave the same bandwidth when I apply Biweight, Triangle and Epan2 kernel function with the default cross validation bandwidth selection.
The following is my code and output
Code:
npregress kernel epsilon ptexp afqt, kernel(triangle) estimator(constant) noderivatives
Code:
npregress kernel epsilon ptexp afqt, kernel(biweight) estimator(constant) noderivatives
Code:
npregress kernel epsilon ptexp afqt, kernel(epan2) estimator(constant) noderivatives
In my regression, ptexp is potential experience, it is an integer variable which can take value from 1 to 20. afqt is an standardized test score which is continuous.
I use the exactly the same data set for each nonparametric regression. The only difference is the choice of kernel function.
The big issue here is that the bandwidth for ptexp is 4.004203 and for afqt is 0.7226754 all the time no matter which kernel function I use.
I don't understand why I got the same bandwidths even if I chose different kernel functions. Is it possible that I may get the same bandwidth no matter which kernel function I used?
Thank you so much
Xizi
