In my earlier plea for advice/suggestions (reproduced below), I should have
said in the penultimate paragraph: "MINITAB attempts to fit a model in
which the participant effect has n - 1 d.f. rather than n -2" (I had n = 17
in mind when I wrote!).
Eddie Shoesmith
*****************
Could anyone give advice on setting up the following 'repeated-measures'
ANCOVA model in MINITAB? It's quite straightforward to specify in SPSS, but
I haven't yet found the appropriate way in MINITAB.
I have a response variable Y that has been measured twice on each of n
participants, under two different conditions of a factor F. A covariate X
has also been measured on each of the n participants. An ANCOVA is required
that includes F and X (and their interaction), and also takes account of the
'repeated-measures' nature of the Y measurements (i.e. takes account of
participant effects).
The model GLM F + X + F*X + P, where P codes for the n participants,
doesn't work because MINITAB detects a collinearity/redundancy between the
participant coding and the covariate measurements, i.e. MINITAB attempts to
fit a model in which the participant effect has 16 d.f. rather than 15.
There's obviously something I'm missing out or not getting quite right. Any
suggestions, please?
Eddie Shoesmith
|