Re: type 3 patches

[Jason Stover]

On Mon, Jan 17, 2011 at 01:27:29PM +0000, John Darrington wrote:
> My initial reaction is to prefer the first patch.  It seems simpler.
> 
> However I'm a bit confused about this :
> 
>      +      gsl_vector_set (result, dropped, 
>      +                      gsl_matrix_get (small_cov, 0, 0)
>      +                      - gsl_matrix_get (small_cov, 0, 0));
> 
> Doesn't that mean that the result will always be the zero vector?

Yes, it should be gsl_matrix_get (small_cov, 0, 0) - sse. I'll test
it before committing.

> To answer the question about in which file to put the function, I suggest
> that for now we put it in glm.c  That's where we know it's going to be used.
> We can move it later if necessary.

OK, I'll start on it.

-Jason