Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: Kernel regressions


From   "Nick Cox" <[email protected]>
To   <[email protected]>
Subject   st: RE: Kernel regressions
Date   Tue, 14 Feb 2006 12:46:55 -0000

This, and related questions, have often been 
discussed on Statalist and I continue to get 
private emails about such questions 
for reasons that will become clear. 

Here is my version of the facts. 

A key principle here is that authors
are to be considered responsible for and 
knowledgeable about their programs, rather than 
Statalist collectively. 

0. -kernreg- was published by the original 
authors in STB-30. 

1. -kernreg1- was published by Xavi Ramos
via Statalist as a partial fix of 
various bugs in the original. 

2. -kernreg2- was published by me as
a fuller fix, ditto. It does not 
have a separate help file, as it just 
does what -kernreg- claims to do, 
or so I believe. 

As I recall, there were emails on 
Statalist of the form "-kernreg- doesn't 
work", which is why people watching the 
list posted 1 and then 2. The original 
authors are not, or have not remained, 
members of Statalist. 

The normal etiquette, and here I put 
on a hat as an Editor of the Stata Journal, 
which superseded the STB in 2001, 
would be for the original authors to re-release
-kernreg- with bug fixes, so that -kernreg2- 
at least can then be zapped as irrelevant. (If
this happened, we would try to contact Xavi Ramos, 
the author of -kernreg1-, and ask him to remove it 
from SSC. I don't know without checking if he 
is on the list.) 

But despite several public and private requests
to do this the original authors have never got round
to it. 

However, and to cut this long story short, 
I consider -kernreg- to be part
of history. -locpoly- is a much better and
more up-to-date program. 

(So, why then does (e.g.) -kernreg2- remain visible? 
Because, whether or not I like it, people still
insist on downloading -kernreg- and using it. 
If they are on Stata <= 7, -locpoly- is no use
to them. At least -kernreg2- remains visible to those
who look carefully.) 

In short, this is a small mess, but you can 
cut straight through and use -locpoly- instead. 

Incidentally, a simple 

. search kernel regression

would have pointed directly at -locpoly-. 

I'd add a not-so-subtle pointer to the 
different but not completely different -mrunning- 
which can handle multiple predictors. 

Nick
[email protected] 

A Briggs
 
> I want to run a kernel regression of y on x with the ultimate 
> aim to plot 
> the smoothed y on x. Searching for user-written programs I have found 
> kernreg, kernreg1 and kernreg2 which are exactly the commands I want.
> 
> However, it appears that these commands are fairly old and graph the 
> smoothed y on x using pre-stata 8 graph styles. I can easily 
> get around 
> this by storing the smoothed variable and the grid-points and 
> then plotting 
> those variables in stata 8 but I was wondering if this 
> program has been 
> updated recently so I can run what I want in one line and not have to 
> generate new variables?

*
*   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/



© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index