Dear K and Manfred,
The easiest option would probably to proceed in 2 steps:
- at the subject level calculate a contrast with the treatment effect
you're interested in. For example cond_post minus cond_pre. If the
interval is not the same for all your subjects and you think this delay
difference could affect the treatment effect, you actually could divide
the individual pre-post difference by that interval duration. This would
give you a "rate of treatment effect" (assuming a linear effect of time).
- bring these contrasts (1 per subject) to the 2nd level and 2 a
2-sample t-test to compare the 2 groups, with unequal and independent
variance.
HTH,
Chris
Le 13/12/2017 à 09:13, Manfred Klöbl a écrit :
> Dear Beginner K.,
>
> regarding your sample size, a repeated-measures model (flexible
> factorial) should be the right choice. According to your description,
> you will have a group, a time and a subject (repeated) factor. Groups
> and subjects are independent, the time factor, however, is not. Each
> subject will appear at each time point, making your post-intervention
> measurements dependent on the pre ones (also also their error
> estimates). When it comes to variance, it's not so easy. There are
> recommendations like keeping the variance setting to "equal" except
> for between-group factors ("group" in your case) but this actually
> depends on your data.
>
> Best regards,
> Manfred
>
>
> Am 13.12.2017 um 05:07 schrieb 김민경:
>> Dear SPM community,
>>
>> Hello, I’d like to ask you for advice.
>>
>> My experiment design is consist of two groups, and each group has 10
>> subjects. Each participant performed fmri twice(pre vs post
>> intervention), and I want to compare the group difference according
>> to repeated measurements.
>> I am worried about what kind of analysis should be done for this.
>> Flexible factorial design with group(variance: unequal,
>> independence), subject,(variance: unequal, independence),
>> time(variance: unequal, independence) factor would be appropriate?
>>
>> Waiting for your reply… thank you.
>>
>> Beginner K.
>
|