[ADMB Users] Problems estimating variance component(s)

[Chris Gast]

Thanks, Dave.  The version that I posted is the "working" version.  The
non-working version can be created by simply changing

 totL += (.5)*(-(nyears-1)*log(tau)-.5*norm2(e/tau));

to

 totL += (-(nyears-1)*log(tau)-.5*norm2(e/tau));

near the end of the procedure section.



Chris




-----------------------------
Chris Gast
cmgast at gmail.com


On Sat, May 1, 2010 at 1:25 PM, dave fournier <otter at otter-rsch.com> wrote:

>
> when I fit the model I get
>
>
> # Number of parameters = 17  Objective function value = 85.0721  Maximum
> gradient component = 1.39976e-05
> # logparms:
>  4.60684 4.96995 5.68780 5.52954 5.57271 5.50498 5.41264 5.38455 5.15434
> 4.90537 5.12176 4.88580 4.99652 4.99314
> # beta:
> 0.323181499499
> # logcmu:
> -1.39791821040
> # logtau:
> -1.02730287171
> # e:
>  -0.135045466851 -0.150301378413 -0.00345438922902 0.258085056965
> -0.0883788436773 -0.207740221584 -0.0111395937873 -0.00399462178862
> 0.0798834043456 -0.0588602853083 -0.00631965917273
> ~
> Is that supposed to be bad?
>
> ~
> _______________________________________________
> 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/20100501/31210b96/attachment.html>