 |
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: pseudo R2s for Generalized Linear Models
On 19 Jul 2007, at 17:55, n j cox wrote:
On the whole, I regard the brandishing of these
single figures of merit as an unsavoury macho exercise,
along the lines of "my model is really good
(considering)!" or (in some weak sciences)
"my model is pretty lousy (but not nearly
as lousy as it could be)!".
Another point to consider is that the catchphrase "percentage of
variance explained" assumes that the percentage of variance
explainable is 100%.
This only happens when there is no measurement error in the predicted
variable. Furthermore, measurement error in the predictors being
considered in the model will reduce this figure still further.
I have had the amusing experience of having data in which doctors
wrote down the persons height and weight in metric and imperial
measures. The percentage of variance in the metric measurements
explained by the imperial measurements was far from 100%!
I would rather see effect sizes and their confidence intervals. This
gives me some idea of the practical significance of the model.
P Before printing, think about the environment
=================================
Ronan Conroy
[email protected]
Royal College of Surgeons in Ireland
120 St Stephen's Green, Dublin 2, Ireland
+353 (0)1 402 2431
+353 (0)87 799 97 95
*
* 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/