04:09 PM
Subject:
Re: [R] Higher log-likelihood in null vs. fitted model
Sent by:
r-help-boun...@r-project.org
Interesting. If you use the deviances printed out by the fitted model
(including the null deviance) and back-engineer it to log-likelihoods, you
get results identical to Stata.
>
Interesting. If you use the deviances printed out by the fitted model
(including the null deviance) and back-engineer it to log-likelihoods, you
get results identical to Stata.
> m$deviance*(-.5)
[1] -390.9304
> m$null.deviance*(-.5)
[1] -393.064
However, using the deviance to calculate the AI
Hi Duncan: I don't know if the following can help but I checked the code
and logLik defines the log likelihood as (p - glmobject$aic/2) where p is
the glmobject$rank. So,
the reason for the likelihood being less is that, in the null, it ends up
being ( 1 - glmobject$aic/2) and in the other one i
On 12-05-31 8:53 AM, Andrew Miles wrote:
Two related questions.
First, I am fitting a model with a single predictor, and then a null model
with only the intercept. In theory, the fitted model should have a higher
log-likelihood than the null model, but that does not happen. See the
output belo
Two related questions.
First, I am fitting a model with a single predictor, and then a null model
with only the intercept. In theory, the fitted model should have a higher
log-likelihood than the null model, but that does not happen. See the
output below. My first question is, how can this happ
5 matches
Mail list logo
