My opinion is that this would miss the point.
It seems an indirect way of comparing smoothers
to me. Degree and kind of smoothing are difficult to separate
in this way, quite apart from working out how many df
you have here.
Given one predictor, one response and two smoothers,
the key information is given by a graph, and you
can bring all you know about the problem to bear on that
(underlying science; purpose of analysis?). Why
throw all this into one omnibus test?
Nick
[email protected]
[email protected]
>
> I apologize, the book is indeed by Fan, and Gijbels. However,
> my question still
> stands for the authors of -locpoly-. Whether the references
> listed below are
> sufficient in explaining the -locpoly- algorithm ?
>
> Concerning the statistical test, what is your opinion on
> testing the distance
> between the distribution of the parametric "y-hat" and the
> distribution of the
> non-parametric "conditional-y-mean" ? Such as the original
> Kolmogorov-Smirnov
> or Anderson-Darling tests.
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
