We found the difference between two levels of factor “a” in our 3dANOVA2. The set up was, 2*8 (hits/misses * subjects). We let the ANOVA run the difference as opposed to running a contrast for each individual subject and then input the glted coefficients (area under the curve from lag 2-4) into the ANOVA. The two levels of factor “a” were hits (when someone got a right answer), and miss (when someone got a wrong answer) on visually presented event related stimuli.
Basically what we want are the areas of activation when someone gets a hit, excluding areas of activation when someone gets a miss.
In looking at our ANOVA contrast we found that practically none of the areas of activation found in either the mean hit minus baseline or mean miss minus baseline corresponded to the areas of activation in our hit minus miss contrast. This was disconcerting and did not make sense to us. How could we subtract something from another thing and get output that is not spatially associated with either of the inputs?
Are we looking at it wrong?
If we do have a problem, then does it involve the number of degrees of freedom we are using? Might it be better to use a region of interest mask from our hits-baseline and miss-baseline ANOVA’s, and then find the difference between the two using some other afni program, like a t-test comparison?
Or could we look at it another way?
What we want, is a contrast where areas of activation in our hits-baseline, that are highly statistically significant (say yellow on the afni statistic color scale), will be seen in our contrast if our misses-baseline has less statistically significant activation in that area (say dark red on the afni statistic color scale). So is it a difference of the statistics that we want, not necessarily the coefficients?
If so then what might be a good way to go about obtaining such a contrast?
Any response would be greatly appreciated and I would be happy to give further clarification if the depths of my incompetence far out reach the current expression of my problem.
Thank you much,
Jeremy