Hello again,
Sorry it has taken me a while to reply. To remind you, the question I had asked was concerning
running a group comparison with a covariate of non-interest (age). When I run the second design
below, which you said was the correct design, I get the following error message,
"Problem with processing model: Warning - design matrix uses different groups (for different
variances), but these do not contain separable " EVs for the different groups (it is necessary that,
for each EV, only one of the groups has non-zero values)."
Do you still think the second of the two designs below, with the age EV as one column and 1s and
2s in the group column, is the correct design? If so, is it okay to ignore this error message and
trust the results of this analysis? Thanks.
Liz Reynolds
On Fri, 18 May 2007 06:50:18 +0100, Steve Smith <[log in to unmask]> wrote:
>Hi,
>
>
>On 17 May 2007, at 17:25, Elizabeth Reynolds wrote:
>
>> Hello all,
>>
>> I am comparing two groups performing a cognitive task. The two groups
>> differ significantly in age but this is unintentional so I want to use
>> age as a covariate to ask the statistical question, "if these two
>> groups were the same age, what activation pattern would I see?". I
>> have seen posts suggesting that I run this analysis in FSL with age as
>> a separate EV for each group where the ages are demeaned from each
>> groups mean separately. See below. This model is simplified as I
>> actually
>> have 9 subjects in each group.
>>
>> Group EV1(group1) EV2 (group2) EV3 (age group1) EV4 (age group2)
>> 1 1 0 4 0
>> 1 1 0 -2 0
>> 1 1 0 -3.5 0
>> 1 1 0 1 0
>> 2 0 1 0 -.75
>> 2 0 1 0 3
>> 2 0 1 0 -2
>> 2 0 1 0 2
>>
>> My concern with this method of analysis is that it is not accounting
>> for differences that may exist between my groups. Instead I think
>> that I should to set up my analysis as below with one EV for age,
>> where ages are demeaned from the grand mean age of the two groups
>> combined.
>
>Indeed, that is correct - you need the second design.
>
>The limitation of this correction is that it assumes that the
>confounding effect of age is linear, but that may be ok.
>
>> When I do this I get an error from FSL, though the analysis
>> runs and produces viable looking results.
>
>What is the error?
>
>> Is the method below the
>> proper way to set up an analysis to ask the question, "if these two
>> groups did not differ by the covariate, what would the activation look
>> like?" and if so is it safe to trust the results FSL is giving me
>> despite the error message? I believe I could get rid of the error
>> message
>> by modeling my analysis as if all subjects came from the same group
>> (having
>> all 1s in the left hand column), however, this does not seem
>> appropriate
>> especially since one group has a genetic disorder and the other
>> does not.
>> Please let me know what you think.Thanks.
>>
>> Group EV1(group1) EV2 (group2) EV3 (age )
>> 1 1 0 3
>> 1 1 0 1
>> 1 1 0 -2.5
>> 1 1 0 4
>> 2 0 1 -1
>> 2 0 1 2
>> 2 0 1 1.7
>> 2 0 1 1.5
>>
>> Liz Reynolds
>
>
>------------------------------------------------------------------------
>---
>Stephen M. Smith, Professor of Biomedical Engineering
>Associate Director, Oxford University FMRIB Centre
>
>FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
>+44 (0) 1865 222726 (fax 222717)
>[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
>------------------------------------------------------------------------
>---
>===========================================================
==============
|