Re: [Marsbar-users] Significant differences for a given ROI

[Matthew B. <mat...@uc...>]

Hi,

> I have defined several ROIs on the basis of individual anatomical
> data. Since I have 11 subjects, I want to know the difference in
> activity between two conditions for each ROI for each of the subjects
> and then do a Random Effect on these data. Now I have two ways to do
> this : 
> 1. I let marsbar, for each subject, determine de difference in beta
> for these two conditions and put this difference into a statistical
> test (t-test relative to 0), or 
> 2. I let marsbar calculate the beta values for each of these
> conditions, and put these beta values into a statistical test (t-test
> for 2 dependent samples).

Er... I think for 1) you are doing a contrast for condition 1 -
condition 2 or somesuch, across all sessions.  Then you take the effect
size values for this, for each subject, and compare them to 0.  This
seems to me the most straightforward way to do it - sad to say!

I'm surprised at the difference between the two approaches though.  Just
as a reality check, if you use the values from approach 2 and calculate
the condition 1 - condition 2 per subject, 2, you should get the same
answer (for subject 1, subject 2 etc) as for approach 1...

See you,

Matthew