| |
 |
|
|
Science Forum Index » Statistics - Education Forum » Binary Logistic Regression Lack of Fit
Page 1 of 1
|
| Author |
Message |
| Bob |
 Posted: Thu Mar 20, 2008 10:05 am |
|
|
|
Guest
|
What is the impact of a significant (p < .05) Hosmer-Lemeshow
goodness of fit test on the parameter estimates of a binary logistic
regression model? Thanks! |
|
|
| Back to top |
|
| Paul Rubin |
| Posted: Thu Mar 20, 2008 8:43 pm |
|
|
|
Guest
|
Bob wrote:
Quote: What is the impact of a significant (p < .05) Hosmer-Lemeshow
goodness of fit test on the parameter estimates of a binary logistic
regression model? Thanks!
I'm not sure it's appropriate to think in terms of the parameter
estimates. Rejection of the null hypothesis indicates that the binary
logistic model is inappropriate (subject to the usual qualifier about
Type I risk). I suppose that might be due to estimation error, but my
first inclination would be to interpret it as indicating model
misspecification -- wrong combination of variables, wrong functional
form, maybe wrong link function.
/Paul |
|
|
| Back to top |
|
| Frank E Harrell Jr |
| Posted: Fri Mar 21, 2008 10:53 am |
|
|
|
Guest
|
Paul Rubin wrote:
Quote: Bob wrote:
What is the impact of a significant (p < .05) Hosmer-Lemeshow
goodness of fit test on the parameter estimates of a binary logistic
regression model? Thanks!
I'm not sure it's appropriate to think in terms of the parameter
estimates. Rejection of the null hypothesis indicates that the binary
logistic model is inappropriate (subject to the usual qualifier about
Type I risk). I suppose that might be due to estimation error, but my
first inclination would be to interpret it as indicating model
misspecification -- wrong combination of variables, wrong functional
form, maybe wrong link function.
/Paul
This test is now obsolete and should be replaced with guided tests of
departures from the model (relaxing the linearity or additivity
assumptions) or by the following which is implemented in the R and
S-Plus Design package's residuals.lrm function -Frank Harrell
@ARTICLE{hos97com,
author = {Hosmer, D. W. and Hosmer, T. and {le Cessie}, S. and
Lemeshow, S.},
year = 1997,
title = {A comparison of goodness-of-fit tests for the logistic
regression
model},
journal = Stat in Med,
volume = 16,
pages = {965-980},
annote = {goodness-of-fit for binary logistic model;difficulty with
Hosmer-Lemeshow statistic being dependent on how groups are
defined;sum of squares test;cumulative sum test;invalidity
of naive
test based on deviance;goodness-of-link function;simulation
setup} |
|
|
| Back to top |
|
| Paul Rubin |
| Posted: Fri Mar 21, 2008 7:56 pm |
|
|
|
Guest
|
Frank E Harrell Jr wrote:
Quote: Paul Rubin wrote:
Bob wrote:
What is the impact of a significant (p < .05) Hosmer-Lemeshow
goodness of fit test on the parameter estimates of a binary logistic
regression model? Thanks!
I'm not sure it's appropriate to think in terms of the parameter
estimates. Rejection of the null hypothesis indicates that the binary
logistic model is inappropriate (subject to the usual qualifier about
Type I risk). I suppose that might be due to estimation error, but my
first inclination would be to interpret it as indicating model
misspecification -- wrong combination of variables, wrong functional
form, maybe wrong link function.
/Paul
This test is now obsolete and should be replaced with guided tests of
departures from the model (relaxing the linearity or additivity
assumptions) or by the following which is implemented in the R and
S-Plus Design package's residuals.lrm function -Frank Harrell
@ARTICLE{hos97com,
author = {Hosmer, D. W. and Hosmer, T. and {le Cessie}, S. and
Lemeshow, S.},
year = 1997,
title = {A comparison of goodness-of-fit tests for the logistic
regression
model},
journal = Stat in Med,
volume = 16,
pages = {965-980},
annote = {goodness-of-fit for binary logistic model;difficulty with
Hosmer-Lemeshow statistic being dependent on how groups are
defined;sum of squares test;cumulative sum test;invalidity of
naive
test based on deviance;goodness-of-link function;simulation
setup}
Thanks for the citation.
/Paul |
|
|
| Back to top |
|
| |
|
Page 1 of 1
All times are GMT - 5 Hours
The time now is Sat Sep 06, 2008 12:14 am
|
|