Hi Jorge,
I'll assume your talking about vbm or tbss group study...
> 1- I want to introduce in my matrix some categorical
> covariables (like gender, smoker-no_smoker..) to remove its
> effects from the results when comparing two groups (controls
> and patients), and I do not know how to do it. I put two
> columns for each covariable (for example: control_gender and
> patients_gender)
You should only do so if you expect an interaction between group and gender. If not, then you should put the gender for both groups in the same column (say 0 for men and 1 for women, and then demean for the entire sample).
> In the control_gender column the patients
> will always hace a zero, but: What should I write for the
> controls? 1 for man and 2 for woman? 1 for man and -1 for
> woman? How can I make the software understand that these
> covariables are categorical and not quantitative?
No need for that... If there is an interaction, then demean the values for the control group only in the control_gender column etc.
> 2-My second question is about the demean process. I have
> found significative differences just when doing demean in my
> quantitative covariables and I don not know how right it
> is.. Should I do the demean even when the covariable does
> not have a normal distribution?
I am not sure I understand your question here... Again, if doing a vbm or tbss study, you need to demean your variables, then run randomise with the -D option if you have not modelled the mean for each group. If you have, then no need to add the -D option. Anything found significant this way after correction for multiple comparisons is valid.
Cheers,
Gwenaelle
--------------------------------------------------------------------
Gwenaëlle Douaud, PhD
FMRIB Centre, University of Oxford
John Radcliffe Hospital, Headington OX3 9DU Oxford UK
Tel: +44 (0) 1865 222 523 Fax: +44 (0) 1865 222 717
www.fmrib.ox.ac.uk/~douaud
--------------------------------------------------------------------
|