Dear Yoram,
> If I use these covariates will the contrast [1 -1] for cases in which
> the counts are higher during the traumatic event repetition than during
> the baseline, but the counts are decreasing from one repetition to the
> next?
No. There are two hyotheses here; (i) the counts are higher during the
traumatic event repetition than during the baseline and (ii) counts are
decreasing from one repetition to the next. As such, these hypotheses
are tested jointly with a conjunction of [1 0] and [0 -1]. The average
of the two (i.e. [1 -1]) is not the same and difficult to interpet.
> If I want to use these covariates to compare between the groups should I
> use the same covariates but within the Multi group*covariate model?
Yes indeed. This is an effect x group interaction and would use a
contrast like [0 1 0 -1] for the repetition-dependent increase
differential (group 1 > group 2).
With best wishes - Karl
> > > I am conducting a research on the response of PTSD and non PTSD patient
> > > to the repetition of their traumatic event.We have 2 group one of
> > > PTSD patients and one of non PTSD patients.For each patient we have 3
> > > baselines and 4 repetition of the traumatic story.
> > >
> > > We want to check the difference in the reaction to the traumatic script
> > > between the PTSD and control group.
> > >
> > > We use three type of covariates
> > >
> > > constant response - 0 0 0 1 1 1 1
> > > rise during the repetition - 0 0 0 1 2 3 4
> > > decrease during the repetition - 0 0 0 3 2 1 0
> > >
> > > These covariates are not orthogonal. If I want to check in what region
> > > of the brain the response correspond to each of the covariate how
> > > should I define my contrasts?
> >
> > The repetition-dependent increases and decreases are modelling the same
> > effect and you only need to specify one regressor:
> >
> > main effect of traumatic script - -1 -1 -1 1 1 1 1
> > script x [linear] time interaction - 0 0 0 -3 -1 1 3
> >
> > Note that these regressors are orthogonal (and orthogonal to the
> > constant term) and can be tested with contrasts [1 0], [-1 0], [0 1]
> > and [0 -1].The last two give you increases and decreases
> > respectively.
> >
> > I hope this helps - Karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|