[R] Running a likelihood ratio test for a logit model

[Prof Brian Ripley]

On Thu, 18 May 2006, Dimitris Rizopoulos wrote:

> the print method for glm object already shows you the null and
> residual deviance with the associated df; look also at the "Value"
> section of ?glm.

Note though that deviances are not (log) likelihood ratios, but 
differences in deviances are twice log LRT (pace a previous answer).

anova() is a good way to get suitable tests out of model fits:
in this case

     anova(glm(y ~ x1, family=binomial))

shows you the appropriate 2 log LRT

For the record, the null model always includes any offsets, and it 
includes and intercept iff the full model does.


> ----- Original Message -----
> From: "Chris Bergstresser" <chris at subtlety.com>
> To: <r-help at stat.math.ethz.ch>
> Sent: Thursday, May 18, 2006 9:20 AM
> Subject: [R] Running a likelihood ratio test for a logit model
>
>
>> Hi all --
>>
>>   I have to calculate a likelihood ratio test for a logit model.  I
>> found logLik, but I need to calculate the log likelihood for the
>> model
>> without any predictors.  How can I specify this in glm?  If the full
>> model is glm(y ~ x1), is the one without predictors (y ~ 0)?  Or (y
>> ~
>> 1)?
>>   Is there a more direct way of getting this?


-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595