Re: Type 3 sums of squares algorithm

[Jason Stover]

On Sat, Jul 09, 2011 at 08:21:10AM +0000, John Darrington wrote:
> In the function get_ssq, there is the code
> 
>       for (i = 1; i < covariance_dim (cov); i++)
>         {
>           if (vars[i] == cmd->design_vars[k])
>             {
>               dropped[n_dropped++] = i;
>             }
>         }
> 
> If I understand it correctly, this loop starts from i = 1 rather
> than i = 0, because we're interested only in the explanatory variables,
> not the dependent ones.
> 
> That being the case, we should change this to
>       for (i = cmd->n_dep_vars; i < covariance_dim (cov); i++)
> 
> So that it'll work for multivariate analysis sometime in the future?

Sounds good to me.