Hello,
A while ago I posted a question regarding contrasts in parametric designs, here is my experiment and previous question:
- I have a group of subjects that performed a task where the stimuli (various geometric forms) presented had different contrast to background levels and different sizes. I used SPM12 to set up a contrast (1st-level analysis) to look at regions where activation increases with increasing contrast to background (parametric regressor 1, so c1 = 0 1 0) and a contrast to look at regions where activity decreases with increasing size of stimuli (parametric regressor 2, so c2 = 0 0 -1) [the first column corresponding to task onset, the second to parametric regressor 1 and the third to parametric regressor 2].
At the group level now I want to look at regions where activation both increases with increasing contrast to background and decreases with the size of stimuli since I am interested in regions that respond to both these characteristics but in opposite directions.
I had two suggestions to do this at the 2nd-level analysis:
1- To perform a conjunction analysis at the group level, estimate the contrasts [0 1 0] and [0 0 -1] at the first level and bring these two contrasts per subject at the second level in a two-sample t-test. The conjunction would then be tested by selecting both contrasts [1 0] and [0 1].
2- Estimate the contrasts [0 1 0] and [0 0 1] at the first level, do two separate 2nd-level analysis using one-sample t-test with contrasts [1 0] for the activation increases and [0 -1] for activation decreases and then compute the conjunction map using imcalc with the expression i1+i2 > 1.
I have two questions now since I am still struggling a little bit understanding contrasts in SPM.
1- I tried both approaches and the maps obtained are similar (although not exactly the same) but when I extract the parameters (I used marsbar) the values are not the same for the two approaches. So, which approach would be the most correct one and why/where do they differ?
2- My second question is: I also want to compare two groups of participants regarding the two first-level contrasts. For the activity increases I think I can see how to set the contrasts at the 2nd-level (I’d use the [0 1 0] from first-level, bring it to the 2nd-level in a two sample t-test and set contrasts [1 -1] and [-1 1] to compare the groups). However, for the activity decreases I am not sure if I should use the 1st-level contrast [0 0 1] or [0 0 -1]?
Thanks for the help understanding this,
|