[ADMB Users] calculating a median

[Philipp Doebler]

I think what JC is hinting at is, that the median optimizes an L1 loss
function. So a variant of the code can work. For example:

http://stats.stackexchange.com/questions/34613/l1-regression-estimates-median-whereas-l2-regression-estimates-mean


That means by optimising the sum of the absolute values of the observations
minus a parameter which estimate is going to be the median you can obtain
the median. This is is many situations computationally less intensive than
sorting.

Best, Philipp


On Fri, Nov 30, 2012 at 4:21 AM, JC Quiroz <juancarlos.quiroz at ifop.cl>wrote:

> Maybe this will help
>
> # ******** dat file
> # n
> 10
> # y
> 1.4  4.7  5.1  8.3  9.0  14.5  14.0  13.4  19.2  18
>
>
> // ******** tpl file
> // Likelihood (short)
> DATA_SECTION
>   init_int n
>   init_vector y(1,n)
>
> PARAMETER_SECTION
>   init_number mu
>   init_number logSigma
>   objective_function_value f
>
> PROCEDURE_SECTION
>   f=n*logSigma+sum(square(y-mu))/(2.0*square(exp(logSigma)));
>
>
> cheers
> JC
>
> ><((((°> ><((((°> ><((((°> ><((((°> ><((((°>
> JC Quiroz
> División de Investigación Pesquera
> Instituto de Fomento Pesquero (IFOP)
> Blanco 839, Valparaíso
> Chile
> tel: 56-32-2151418
> fax: 56-32-2151645
> email: juancarlos.quiroz at ifop.cl
> web: http://codemorecode.wordpress.com/
>
>
>
> ----- Original Message -----
> From: "Melissa T Drake (DNR)" <melissa.drake at state.mn.us>
> To: users at admb-project.org
> Sent: Thursday, 29 November, 2012 19:03:29 (GMT-0400) Auto-Detected
> Subject: [ADMB Users] calculating a median
>
>  Is there an easy way calculate the median value of a vector?
>
>
>
> Melissa T. Drake
>
> Fisheries Research Supervisor
>
> Minnesota Department of Natural Resources
>
> 500 Lafayette Road
>
> St. Paul, MN 55155
>
> 651-259-5245
>
> Melissa.drake at state.mn.us
>
>
>
> _______________________________________________ Users mailing list
> Users at admb-project.org
> http://lists.admb-project.org/mailman/listinfo/users
>
> _______________________________________________
> Users mailing list
> Users at admb-project.org
> http://lists.admb-project.org/mailman/listinfo/users
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.admb-project.org/pipermail/users/attachments/20121130/5e70003d/attachment.html>