[ADMB Users] SEPARABLE_FUNCTION

[Yuying Zhang]

Hello Dr Fournier,

Thank you very much for your quick reply. Maybe MCMC is an easier way 
for me.

I have another question for you.

As I mentioned in my previous message, I need the sample sizes because I 
want to conduct T tests:
(mean1-mean2)/sqrt(se1*se1/N1+se2*se2/N2)

A friend suggested me to ignore the adjustments of sample sizes. 
According to him, "the standard errors in the std file automatically 
handle having more or less data (also poor or better data).  They are 
literally the standard errors for the estimates in the std file, like 
the standard error of the mean = sigma/sqrt(N).  I think the effects of 
are already in the standard error before it is written to the std file."

I want to confirm with you if his understanding is right?

Yuying

On 3/10/2011 6:57 AM, dave fournier wrote:
> you need to do ALL the calculations inside the separable function.
> Maybe someone can post an example for you.
>
> Otherwise if you have a large number of random effects you could try the
> -ndb N option such as
>
>      -ndb 10    or -ndb 25
>
> the separable function with sparse matrix should be faster however.