[R-sig-ME] Evaluating mixed logistic model efficiency
Drew Tyre
atyre2 at unl.edu
Fri Dec 3 16:50:46 CET 2010
Arnaud,
If you have binary 0/1 data one widely used method for looking at the
quality of the fit uses the Area under the Receiver Operating
Characteristic Curve - variously abbreviated ROC or AUC. It gives the
probability your model will correctly rank any two observations. There
is a nice package PresenceAbsence which does the calculations and
makes pretty plots.
If you have binomial data you can compare the residual deviance with a
Chisquare distribution with n-p df (n is sample size and p is the
number of parameters), but I'm not sure how well that works in the
mixed model case. I look forward to other suggestions.
On Fri, Dec 3, 2010 at 9:20 AM, Arnaud Mosnier <a.mosnier at gmail.com> wrote:
> Dear mixed modelers,
>
> In my actual knowledge status, the best available measures to evaluate
> goodness-of-fit of a mixed logistic model are Deviance, LogLikelihood, AIC,
> AICc if necessary, BIC (I may forgot some of them but they are all linked).
> However, these measures only permit to classify models from the best to the
> worse without knowing if the best model is really efficient in explaining
> the data.
>
> It exists pseudo-R2 calculation based on differences between the likelihood
> of the intersect only model and the likelihood of the model that you wanted
> to evaluate. But use of those methods are generally discouraged.
>
> I would like your opinion about that !
> If you have any suggestion I will be happy to learn from you !
>
> Thanks for your help.
>
> Arnaud
>
> [[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>
--
Drew Tyre
School of Natural Resources
University of Nebraska-Lincoln
416 Hardin Hall, East Campus
3310 Holdrege Street
Lincoln, NE 68583-0974
phone: +1 402 472 4054
fax: +1 402 472 2946
email: atyre2 at unl.edu
http://snr.unl.edu/tyre
http://aminpractice.blogspot.com
http://www.flickr.com/photos/atiretoo
More information about the R-sig-mixed-models
mailing list