Dear Lena,
Most people use the contrast maps, which as you have noted are
difficult to interpret.
We did a multidimensional analysis of similar task designs if you're
interested, which did not require assuming a HDR shape or computing
contrasts. To compare between groups you do ANOVA on the HDR shapes
which are derived using FIR models:
http://www.cnoslab.com/pdfs/TaskRelated-Functional-Connectivity.pdf
http://www.cnoslab.com/pdfs/impairment-in-subcortical.pdf
Best,
Todd
On Wed, Nov 3, 2021 at 9:40 AM Lim, Lena
<[log in to unmask]> wrote:
>
> Dear SPM experts,
>
> We are trying to determine the best way to create contrast maps across two to three clinical groups for BOLD activation during an emotional go/no-go task with four conditions. For each subject, we have the first-level beta maps for each task condition as specified in the GLM, i.e., fear no-go, happy go, happy no-go, fear go. We also have within-subject contrasts for the primary task comparisons for each subject, e.g., fear go vs happy go and happy go vs fear go. If we now want to determine the difference in BOLD activation across groups, would we run one-way ANOVA and/or t-tests using the contrast maps or would we run these statistical tests using the beta maps?
>
>
> For example, if we want to learn whether there is greater activation in one group compared with another for the happy go trials, would we run a t-test using the happy go beta maps or using the happy go-fear go contrast maps? We have previously done both, but when running a t-test with contrast maps, we have sometimes ended up with strange activations that are outside the typical set of brain areas involved in the task. Is this an artifact of the contrast of contrasts, in that we might be seeing “deactivations” contrasted with activations that are appearing as activations in the contrast?
>
>
> Many thanks,
>
> Lena
>
--
******************************
Todd S. Woodward, Ph.D.
Professor
Department of Psychiatry
University of British Columbia
Vancouver, Canada
|