nuria, while you've got many good advices, i have to say that i will agree with the reviewers. if i were you, i probably will fit a ordinal logit model instead of a gaussian model with continous outcome. just my $0.02USD subject to inflation.

On Thu, Mar 5, 2009 at 4:19 PM, nuria <nchapinal@yahoo.com> wrote: > Hi, > > I am analysing a variable (lameness visual score for cows) that has 10 > levels, from 1 to 5, with half points ( 1, 1.5, 2, 2.5.... and so on). > > I have used GLM and MIXED and I have plotted the residuals and the > assumptions of normalitiy and homogeneity of variance are met. > > The reviewer of a journal is telling me that I shouldn't use GLM or > MIXED. I want to argue that, but I first want to make sure I am right. > Is there any reference that say that parametrics stats are robust > enough for ordinal data? > > I do not want to use my variable as a categorical variable, because it > has 10 levels.... > > Thanks so much >

-- =============================== WenSui Liu Acquisition Risk, Chase Blog : statcompute.spaces.live.com

I can calculate the motion of heavenly bodies, but not the madness of people.” -- Isaac Newton ===============================