Many thanks for your reply, Anderson!
Still need to probe a little further, though, i'm afraid.
Put simply, what we want to be able to do is determine if a single data point – a mean condition beta weight from a single subject derived from an average of four runs – is statistically non-zero. I think that is the key – the single subject part.
(1) Can I coax you into commenting specifically on what I have describe as Feat query Approach-1 vs. Feat query Approach-2?
(2) How does the use of featquery differ from what you’ve suggested – to “average the COPEs within the ROI for each subject (with fslstats or fslmeants)”.
(3) Can I ask you specifically about the featquery outputs? …as perhaps this is the answer we are looking for…
We have the following values (for the COPE that reflects condition B > rest):
Stats/pe: mean = 0.3526, stddev = 0.203
Stats/zstat: mean = 2.721, stddev = 1.405
Can we straightforwardly interpret the mean zstat?
For example, if we compute a p-value based on the mean zstat of 2.721 (two-tailed) = 0.0065; does this then suggest that the mean parameter estimate of 0.3526 is statistically non-zero, p < 0.01?
Thanks again for your time, and patience!
I will feel hugely relieved when i finally better my understanding on this! Cheers.
|