Dear Patricia,
The term "demeaning" simply refers to removing the mean value.
That is, taking the original values and subtracting the mean value from
them (to make them zero mean afterwards).
If you are not interested in the whether there is a different correlation
with age in the different groups, but instead just want to "control for age"
then I suggest that you have a single EV for age and make this a zero
mean EV. That is, demean your age values before putting them into
this single EV. So get the set of age values, calculate the mean value,
subtract this from each value and put the results of the subtractions
into the single age EV. I hope this is clear now.
All the best,
Mark
On 5 Aug 2011, at 12:16, Patricia Pires wrote:
> Dear Mark,
>
> thank you very much for your response. However it is for me, difficult to undertand the process that you're describing.
>
> In my sample i want to analize FA values (in Glm and randomise) comparng these 4 groups expecting there will exist less anisotropy in the more severe stages of the disease. However in the statistical analysis for demographic variables, there are significant differences in age. That's why i include the age in my matrix as a covariable value. What i was trying to do is to remove age factor on the results to observe only anisotropy differences due to group factor.
>
> If I understand correctly, the matrix I designed would be to observe (both) group differences + age so this would not be the correct construction. "Demeaning" would be the procedure to get what I want? And how can i demean? I have been searching in JSCMail but i can't find the procedure. Can you please explain how to demean and how to construct the matrix for this?
>
> Thank you very much,
>
> Patricia.
>
> 2011/8/1 Mark Jenkinson <[log in to unmask]>
> Dear Patricia,
>
> This is not what we would recommend for this analysis because
> the age values are not demeaned. As there is some difference
> between the mean age of each group (16.39 16.59 19.07 17.89)
> then this will be driven by some combination of age *and* group
> difference, which is not what you are likely to want. It will also
> affect the group difference contrasts. So I strongly recommend
> that you demean each group.
>
> Also, splitting age between the different groups allows you to fit
> different slopes (amounts of variation of quantity with age) for
> each group. Is that what you want, or do you just want to account
> for age as a covariate of no interest across all groups? Given
> that you want a different mean age in each group then doing
> the latter is difficult, but you must be aware of what you are assuming
> when you are doing the former.
>
> Jeanette Mumford has a very good discussion of the various
> demeaning issues at:
> http://mumford.fmripower.org/mean_centering/
>
> As for your contrast - C1 asks the question of whether the
> average quantity (fMRI effect size, or FA, or whatever you are
> testing) is different from zero for the first group only. If you
> want to know whether groups differ from each other then you
> need to have differential contrasts such as:
> -1 1 0 0 0 0 0 0
> -1 0 1 0 0 0 0 0
> -1 0 0 1 0 0 0 0
> 0 -1 1 0 0 0 0 0
> 0 -1 0 1 0 0 0 0
> 0 0 -1 1 0 0 0 0
> which ask the questions:
> Is the average quantity in group 2 greater than the average quantity in group 1?
> Is the average quantity in group 3 greater than the average quantity in group 1?
> Is the average quantity in group 4 greater than the average quantity in group 1?
> Is the average quantity in group 3 greater than the average quantity in group 2?
> Is the average quantity in group 4 greater than the average quantity in group 2?
> Is the average quantity in group 4 greater than the average quantity in group 3?
>
> Note that I've given the 8 column version, which is not to meant to mean that
> I agree with splitting the age covariate, I'm just trying to make it more similar
> to the ones you provided. This depends on the points I discussed earlier,
> but the contrasts can be converted to the 5 column version by dropping the last
> 3 zeros.
>
> I hope this helps.
> All the best,
> Mark
>
>
>
> On 1 Aug 2011, at 16:05, SUBSCRIBE FSL Patricia Pires wrote:
>
> > Dear experts,
> >
> > i want to analize in Glm 4 different groups (Controls, Mild, Medium and Severe disease).
> > Also, i have included age covariate in EV's because this was significant.
> > This is the matrix:
> >
> > /Matrix
> > 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 6.300000e+01 0.000000e+00 0.000000e+00 0.000000e+00
> > 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 5.900000e+01 0.000000e+00 0.000000e+00 0.000000e+00
> > 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 5.800000e+01 0.000000e+00 0.000000e+00 0.000000e+00
> > 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.400000e+01 0.000000e+00 0.000000e+00 0.000000e+00
> > 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 6.700000e+01 0.000000e+00 0.000000e+00 0.000000e+00
> > 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 6.000000e+01 0.000000e+00 0.000000e+00 0.000000e+00
> > 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.400000e+01 0.000000e+00 0.000000e+00 0.000000e+00
> > 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 6.900000e+01 0.000000e+00 0.000000e+00 0.000000e+00
> > 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 6.400000e+01 0.000000e+00 0.000000e+00 0.000000e+00
> > 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 6.100000e+01 0.000000e+00 0.000000e+00 0.000000e+00
> > 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 6.000000e+01 0.000000e+00 0.000000e+00 0.000000e+00
> > 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.100000e+01 0.000000e+00 0.000000e+00 0.000000e+00
> > 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 6.400000e+01 0.000000e+00 0.000000e+00 0.000000e+00
> > 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.400000e+01 0.000000e+00 0.000000e+00 0.000000e+00
> > 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 6.600000e+01 0.000000e+00 0.000000e+00
> > 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 6.900000e+01 0.000000e+00 0.000000e+00
> > 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 6.800000e+01 0.000000e+00 0.000000e+00
> > 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 6.300000e+01 0.000000e+00 0.000000e+00
> > 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 6.600000e+01 0.000000e+00 0.000000e+00
> > 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 6.300000e+01 0.000000e+00 0.000000e+00
> > 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 6.600000e+01 0.000000e+00 0.000000e+00
> > 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.100000e+01 0.000000e+00 0.000000e+00
> > 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 6.400000e+01 0.000000e+00 0.000000e+00
> > 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 6.200000e+01 0.000000e+00 0.000000e+00
> > 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.000000e+01 0.000000e+00 0.000000e+00
> > 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 6.200000e+01 0.000000e+00 0.000000e+00
> > 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 6.900000e+01 0.000000e+00 0.000000e+00
> > 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.000000e+01 0.000000e+00 0.000000e+00
> > 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 8.300000e+01 0.000000e+00
> > 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.800000e+01 0.000000e+00
> > 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.900000e+01 0.000000e+00
> > 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.800000e+01 0.000000e+00
> > 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.200000e+01 0.000000e+00
> > 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.400000e+01 0.000000e+00
> > 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.800000e+01 0.000000e+00
> > 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.200000e+01 0.000000e+00
> > 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.900000e+01 0.000000e+00
> > 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.500000e+01 0.000000e+00
> > 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.900000e+01 0.000000e+00
> > 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.500000e+01 0.000000e+00
> > 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 6.900000e+01 0.000000e+00
> > 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.700000e+01 0.000000e+00
> > 0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.700000e+01
> > 0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.400000e+01
> > 0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.000000e+01
> > 0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 6.400000e+01
> > 0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.300000e+01
> > 0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 5.900000e+01
> > 0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.800000e+01
> > 0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 6.700000e+01
> > 0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.900000e+01
> > 0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.100000e+01
> > 0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.700000e+01
> > 0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.300000e+01
> > 0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 7.600000e+01
> > 0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 6.400000e+01
> >
> > And the contrast:
> >
> > /Matrix
> > 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00
> > 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00
> > 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00
> > 0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00
> >
> > Matrix and contrast are well built?
> > If there is any difference in contrasts (e.g. C1) does it means this group differs from all others?
> >
> > Thanks very much,
> >
> > Patricia.
> >
>
|