The column of 1's can, in certain circumstances, mean that it does not matter whether you demean the covariates or not, but this is not true for all cases (i.e. for all contrasts). When there is a difference it is better to demean, and therefore it is safer to demean. So I strongly recommend sticking with the demeaned covariate values, and including the column of 1's.
All the best,
On 19 Apr 2013, at 08:05, Leighton Barnden <[log in to unmask]> wrote:
> Thanks Mark,
> A quick follow up: Will not inclusion of the column of 1's effectively remove the covariate means and therefore render de-meaning of the covariates beforehand unnecessary?