Subject: glm and stepAIC selects too many effects

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

Subject: glm and stepAIC selects too many effects

Ravi Varadhan-2
If AIC is giving you a model that is too large, then use BIC (log(n) as the penalty for adding a term in the model).  This will yield a more parsimonious model.  Now, if you ask me which is the better option, I have to refer you to the huge literature on model selection.

Best,

Ravi

        [[alternative HTML version deleted]]

______________________________________________
[hidden email] mailing list -- To UNSUBSCRIBE and more, see
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
Reply | Threaded
Open this post in threaded view
|

Re: Subject: glm and stepAIC selects too many effects

Ravi Varadhan-2
More principled would be to use a lasso-type approach, which combines selection and estimation in one fell swoop!



Ravi

________________________________
From: Ravi Varadhan
Sent: Tuesday, June 6, 2017 10:16 AM
To: [hidden email]
Subject: Subject: [R] glm and stepAIC selects too many effects


If AIC is giving you a model that is too large, then use BIC (log(n) as the penalty for adding a term in the model).  This will yield a more parsimonious model.  Now, if you ask me which is the better option, I have to refer you to the huge literature on model selection.

Best,

Ravi

        [[alternative HTML version deleted]]

______________________________________________
[hidden email] mailing list -- To UNSUBSCRIBE and more, see
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
Reply | Threaded
Open this post in threaded view
|

Re: Subject: glm and stepAIC selects too many effects

Bert Gunter-2
See package "glmnet".

-- Bert


Bert Gunter

"The trouble with having an open mind is that people keep coming along
and sticking things into it."
-- Opus (aka Berkeley Breathed in his "Bloom County" comic strip )


On Tue, Jun 6, 2017 at 8:10 AM, Ravi Varadhan <[hidden email]> wrote:

> More principled would be to use a lasso-type approach, which combines selection and estimation in one fell swoop!
>
>
>
> Ravi
>
> ________________________________
> From: Ravi Varadhan
> Sent: Tuesday, June 6, 2017 10:16 AM
> To: [hidden email]
> Subject: Subject: [R] glm and stepAIC selects too many effects
>
>
> If AIC is giving you a model that is too large, then use BIC (log(n) as the penalty for adding a term in the model).  This will yield a more parsimonious model.  Now, if you ask me which is the better option, I have to refer you to the huge literature on model selection.
>
> Best,
>
> Ravi
>
>         [[alternative HTML version deleted]]
>
> ______________________________________________
> [hidden email] mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

______________________________________________
[hidden email] mailing list -- To UNSUBSCRIBE and more, see
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.