Each beta in the model has a corresponding t-stat. In AFNI what would be the best way of going about obtaining a combined t stat for two predictors (like contrasting both predictors against the model’s baseline)? Thanks in advance for your advice.
Suppose that the labels for the two predictors are P1 and P2, add the following in your 3dDeconvolve script:
-gltsym ‘SYM: +P1 \ -P2’
The F-statistic in the output will provide you the statistic evidence for the combined effect of the two predictors.
Thanks Gang
Also, could I simply use 3dcalc (.5P1 + .5P2) and then use 3dtest to single sample t test (determine if it differs from zero)? If I am totally off base please fill me in.
could I simply use 3dcalc (.5P1 + .5P2) and then use 3dtest to single sample t test (determine if it differs from zero)?
Are you referring to group analysis instead of individual level? With (.5P1 + .5P2), you are making inference about the average effect between the two conditions.
I am not talking about group analysis. I am working with a single individual and I want to find out how two conditions differ from baseline (the combined effect which is why the average made sense). The goal is to get a t map for this so if I apply 3dcalc to the coeffs for the two conditions how would I get a t map for the combined effect? Thanks Again
If the average effect of the two conditions is what you’re looking for, do this:
-gltsym ‘SYM: +0.5P1 +0.5P2’
Okay thanks again Gang but there is still something I am missing here. What if I wanted to get the general effect across three instead of two conditions (P1, P2, P3) against baseline?
First, correction. My previous message should be:
If the average effect of the two conditions is what you’re looking for, do this:
-gltsym ‘SYM: +0.5P1 +0.5P2’
What if I wanted to get the general effect across three instead of two conditions (P1, P2, P3) against baseline?
It’s still very opaque to me what exactly you mean by “combined” or “general” effect among two or three conditions. Again, if you’re looking for the average effect with three conditions, try
-gltsym ‘SYM: +0.3333P1 +0.3333P2 +0.3333P3’
Okay! Now it makes sense the last answer totally confused me (I was expecting what you wrote in this message) and that is why I asked the follow up question. I should be more clear the average is my objective.
Thanks Again
I read this thread, and this is very similar for something I am looking for.
Assuming I want the contrast between two predictors on the group level (not between two levels of a predictor, which I can get using gltCode in e.g. 3dMVM). How do I do that?
I should also mention that I want a statistical map of this contrast, not just a mask. So 3dcalc is not what I’m looking for. In addition, these two predictors are continuous, so it is actually a contrast of slopes.
Thank you!