Hi Jeanette,
I'm having a problem similar to Hyo's, and your answer was quite clear. Just
to make sure I have a handle on the idea of making the matrix separable,
though, here's my design matrix:
Let ev1=pat, ev2=con, ev3=malepat, ev4=malecon, ev5=femalepat, ev6=femalecon
grp ev1 ev2 ev3 ev4 ev5 ev6
1 1 0 1 0 0 0
1 1 0 0 0 1 0
1 1 0 1 0 0 0
2 0 1 0 0 0 1
2 0 1 0 1 0 0
2 0 1 0 0 0 1
2 0 1 0 1 0 0
For contrasts,
Patient-Control : 1 -1 0 0 0 0
Control: 0 1 0 0 0 0
Patient: 1 0 0 0 0 0
Male-Female: 0 0 1 1 -1 -1
Male: 0 0 1 1 0 0
Female: 0 0 0 0 1 1
How does this look?
Much appreciated,
Laura
On Wed, 5 Nov 2008 16:53:01 -0800, Jeanette Mumford <[log in to unmask]> wrote:
>Hi Hyo
>
>You do need to make a few adjustments to the model. First, ev3 is not
>necessary and with it your design is rank deficient since ev1+ev2=ev3.
>Also, if you would like to compare reaction time between groups you will
>want to separate ev4 into two groups. First demean, as you have done, and
>then separate. Last thing, which won't be an issue when you make the above
>changes, is that when using the 'group' column to indicate different groups
>in your analysis, the model has to be separable. This simply means that if
>an ev has a nonzero entry for one group it must have all zero entries for
>the other groups.
>
>Hope that helps.
>
>Jeanette
>
>On Wed, Nov 5, 2008 at 9:15 AM, Hyo Jong <[log in to unmask]> wrote:
>
>> Hello Experts,
>> I have a question about design matrix for two-sample T-est with a
>> covariate.
>> Let say I have two groups of subjects (eg 3 patients and 4 controls -
>> actual
>> numbers are larger) and response time as a covariate. Two-sample T-test is
>> very clear because an example is explained in a FSL document. I am not sure
>> how to include response time to find out the effect of it.
>>
>> Let ev1=patient, ev2=control, ev3=all, ev4= demeaned response time
>>
>> group ev1 ev2 ev3 ev4
>> 1 1 0 1 24
>> 1 1 0 1 -18
>> 1 1 0 1 -7
>> 2 0 1 1 5
>> 2 0 1 1 -4
>> 2 0 1 1 6
>> 2 0 1 1 6
>>
>> For contrasts,
>> Patient-Control : 1 -1 0 0
>> Control-Patient : -1 1 0 0
>> Response time: 0 0 0 1
>>
>> Are these correct?
>> I am not sure if I should include ev3 and if I should set ev4 for each
>> group to
>> find out response time effect for patients and controls, respectively.
>> Any help will be greatly appreciated.
>>
>> Hyo Jong
>>
>
|