Quantcast

negative value for AIC and BIC

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

negative value for AIC and BIC

Olivier MARTIN
Hi all,


I obtained negative values for AIC and BIC criteria for a particular
model that I have
developped...

I don't remember to have negative values for these crietria for others
applications, so I am a
little suprised... Could anyone tell me if something is wrong or his
conclusion concerning my model?

Best regards,
Olivier.

______________________________________________
[hidden email] mailing list
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
|  
Report Content as Inappropriate

Re: negative value for AIC and BIC

Hannu Kahra
Olivier,

type ?AIC and have a look at the description

Description:

     Generic function calculating the Akaike information criterion for
     one or several fitted model objects for which a log-likelihood
     value can be obtained, according to the formula -2*log-likelihood
     + k*npar, where npar represents the number of parameters in the
     fitted model, and k = 2 for the usual AIC, or k = log(n) (n the
     number of observations) for the so-called BIC or SBC (Schwarz's
     Bayesian criterion).

AIC = -2*log-likelihood + k*npar can be negative as SBC, too.

Hannu

On 9/7/07, Olivier MARTIN <[hidden email]> wrote:

>
> Hi all,
>
>
> I obtained negative values for AIC and BIC criteria for a particular
> model that I have
> developped...
>
> I don't remember to have negative values for these crietria for others
> applications, so I am a
> little suprised... Could anyone tell me if something is wrong or his
> conclusion concerning my model?
>
> Best regards,
> Olivier.
>
> ______________________________________________
> [hidden email] mailing list
> 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.
>

        [[alternative HTML version deleted]]

______________________________________________
[hidden email] mailing list
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
|  
Report Content as Inappropriate

Re: negative value for AIC and BIC

John Sorkin
Oliver,
I am attaching an HTML document in which I have plotted -2Log(x) vs. x. If you examine the plot you will see that -2Log(x) can be negative. Since -2Log(x) is part of AIC and BIC, AIC and BIC can be negative.
John

John Sorkin M.D., Ph.D.
Chief, Biostatistics and Informatics
Baltimore VA Medical Center GRECC,
University of Maryland School of Medicine Claude D. Pepper OAIC,
University of Maryland Clinical Nutrition Research Unit, and
Baltimore VA Center Stroke of Excellence

University of Maryland School of Medicine
Division of Gerontology
Baltimore VA Medical Center
10 North Greene Street
GRECC (BT/18/GR)
Baltimore, MD 21201-1524

(Phone) 410-605-7119
(Fax) 410-605-7913 (Please call phone number above prior to faxing)
[hidden email]

>>> "Hannu Kahra" <[hidden email]> 09/07/07 4:32 AM >>>
Olivier,

type ?AIC and have a look at the description

Description:

     Generic function calculating the Akaike information criterion for
     one or several fitted model objects for which a log-likelihood
     value can be obtained, according to the formula -2*log-likelihood
     + k*npar, where npar represents the number of parameters in the
     fitted model, and k = 2 for the usual AIC, or k = log(n) (n the
     number of observations) for the so-called BIC or SBC (Schwarz's
     Bayesian criterion).

AIC = -2*log-likelihood + k*npar can be negative as SBC, too.

Hannu

On 9/7/07, Olivier MARTIN <[hidden email]> wrote:

>
> Hi all,
>
>
> I obtained negative values for AIC and BIC criteria for a particular
> model that I have
> developped...
>
> I don't remember to have negative values for these crietria for others
> applications, so I am a
> little suprised... Could anyone tell me if something is wrong or his
> conclusion concerning my model?
>
> Best regards,
> Olivier.
>
> ______________________________________________
> [hidden email] mailing list
> 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.
>
        [[alternative HTML version deleted]]

______________________________________________
[hidden email] mailing list
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.

Confidentiality Statement:
This email message, including any attachments, is for the sole use of the intended recipient(s) and may contain confidential and privileged information.  Any unauthorized use, disclosure or distribution is prohibited.  If you are not the intended recipient, please contact the sender by reply email and destroy all copies of the original message.

______________________________________________
[hidden email] mailing list
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.

-2Log.html (442 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: negative value for AIC and BIC

Mike Meredith
In reply to this post by Hannu Kahra
Sure -2*log(x) can be negative, and it can outweigh the k*npar term. Just do:

curve(-2*log(x)+2, 0.1, 10)  # for AIC with npar = 1
abline(h=0, v=exp(1), lty=3)

However, that only happens for x > exp(1) or even bigger if npar > 1. I think Olivier's real question is: do we believe in likelihoods > 1 ?

Cheers, Mike.


Hannu Kahra wrote
Olivier,

type ?AIC and have a look at the description

Description:

     Generic function calculating the Akaike information criterion for
     one or several fitted model objects for which a log-likelihood
     value can be obtained, according to the formula -2*log-likelihood
     + k*npar, where npar represents the number of parameters in the
     fitted model, and k = 2 for the usual AIC, or k = log(n) (n the
     number of observations) for the so-called BIC or SBC (Schwarz's
     Bayesian criterion).

AIC = -2*log-likelihood + k*npar can be negative as SBC, too.

Hannu

On 9/7/07, Olivier MARTIN <olivier.martin@avignon.inra.fr> wrote:
>
> Hi all,
>
>
> I obtained negative values for AIC and BIC criteria for a particular
> model that I have
> developped...
>
> I don't remember to have negative values for these crietria for others
> applications, so I am a
> little suprised... Could anyone tell me if something is wrong or his
> conclusion concerning my model?
>
> Best regards,
> Olivier.
>
> ______________________________________________
> R-help@stat.math.ethz.ch mailing list
> 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.
>

        [[alternative HTML version deleted]]

______________________________________________
R-help@stat.math.ethz.ch mailing list
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.
Loading...