Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: L-statistic


From   n j cox <n.j.cox@durham.ac.uk>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: L-statistic
Date   Fri, 25 Apr 2003 17:40:37 +0100

Joseph Coveney wrote:

As I mentioned in the post yesterday, I couldn't find anything explicit that setting missing-value likelihood residuals to zero is standard practice.

But in fairness to David Collett, I think that he was justified in setting the missing-value residual to zero in this case, if that's what he did: the observed value was 1 and the predicted value was 1; that's a residual of zero.

The hitch in the whole matter arose because the Pearson residual takes the absolute residual and divides by the standard error of the predicted value, which in this case is zero.
This raises an interesting general question. From a computational viewpoint, generating a Pearson residual from

(observed - predicted) / se of predicted

should yield missing whenever the se is 0. But I'd
argue that from a statistical viewpoint there should
be an override that when observed equals predicted, the
Pearson residual should be 0. Hence in more Stata-like language

gen Pearson = (observed - predicted ) / se

should be

gen Pearson = (observed - predicted ) / se
replace Pearson = 0 if observed == predicted

Or perhaps the best of both worlds is that this -replace- should
be available as an option, so that the override must
be conscious.

Nick
n.j.cox







*
* 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 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index