Dear Kenji -
take the gaussian distribution and integrate out the variance (from 0
to infinity) and this form is what you will get.
Best
Tim
On 29 May 2009, at 03:42, kenji wrote:
> Dear FSL experts,
>
> My question is about the log likelihood function.
>
> I saw this by looking at the compute_likelihood.
> m_likelihood_en = (m_data.Nrows()/2)*log(sumsquares/2);
>
> Why are (m_data.Nrows()/2) and log(/2) necessary ?
> Though I think that the form is
> m_likelihood_en = log(sumsquares);
> Maybe I'm missing something?
>
> Thanks for any feedback.
> Kenji
>
|