[ADMB Users] difference between ADMB-RE and R/mgcv in SEs for smoother coefficients in a GAM fitted by maximum likelihood

[dave fournier]

On 12-11-30 06:41 AM, Tim Miller (NOAA Federal) wrote:

Can you post the parameter estimates and the stds for ADMB and the R 
routine. I'm curious to
compare them.


> Yes, Hans is correct. This does fix the covariance issue for my 
> problem. Many thanks to Hans and Dave for sorting this out and so 
> rapidly at that.
>
> Best,
> Tim
>
>
> On 11/29/2012 5:28 PM, H. Skaug wrote:
>> Thanks. That seems to solve the problem, both for
>> my simple example and Tim's original example (but
>> it would be good if Tim could verify  himself).
>>
>> I will submit this to the source base, but until the next
>> release of ADMB is out people must include
>> the attached file in their tpl, using:
>>
>> GLOBALS_SECTION
>>    #include "getbigs.cpp"
>>
>> (I know this is not the best way of doing it, but it
>> is the quickest).
>>
>> Hans
>>
>>
>> On Thu, Nov 29, 2012 at 10:53 PM, dave fournier<davef at otter-rsch.com>  wrote:
>>> OK that helped.  Around line 72 in getbigs.cpp I beleive the sign is wrong.
>>> It should be
>>>
>>>
>>>     uhat_prime=-minv*Dux;
>>>
>>> This corresponds to   uhat'(x) = - (L_uu)^{-1} L_xu
>>>
>>>
>>> _______________________________________________
>>> 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/20121130/29cf99e0/attachment.html>