AFNI version info (afni -ver): AFNI_21.1.04 Hi AFNI experts, I have recently ran a 3dANOVA2 -type3 for within subjects repeated measure comparisons from three time points to assess the treatment effect for fours measures FA, MD, RD, AD. Now , I am wondering, How should I correct for multiple ANOVA's for each measure and later for pothoc testing.
Also , i am not able to find a way to implement cluster correction for Ftest. As 3dANOVA2 do not provide any residuals and residuals are important to get the acf values. Is there any alternative way available to calculate these and then apply 3dClustSim?
Using 3dClustSim to account for spatial smoothness is indeed a common practice in neuroimaging. To obtain the residuals, you can perform your analysis using either 3dMVM or 3dLMEr.
However, typical multiple testing adjustment methods can be excessively conservative due to the questionable assumptions inherent in mass univariate modeling. Instead of focusing solely on stringent thresholding, one may emphasize effect estimation and associated uncertainty. Overly strict thresholds often lead to artificial dichotomization and an increased risk of false negatives. Therefore, it might be unnecessary to make additional adjustments for the four separate analyses related to FA, MD, RD, and AD.
As an alternative approach, you could simply employ a voxel-level p-value threshold (e.g., 0.01) in conjunction with a cluster-level threshold (e.g., 20 voxels). In addition, adopt the highlight-but-don’t-hide strategy, which promotes a nuanced interpretation of neuroimaging results.
Dear Gang,
Thanks for your immediate response. I earlier thought to use 3dLMEr but as my sample size is small only 27 individuals and i thought it might increase the complexity of the model and produce low power for small sample size.
On the other hand, How I can correct for posthoc multiple comparisons. As I have 4 measures and each measure I have three time points to compare. Therefore, the number of posthoc comparisons become 6x4 = 24. Alternatively, Is it possible to correct for post hoc comparisons for each measure i.e., the 6 pairwise comparisons?
I am also wondering that in order to account for spatial smoothness, is it possible to use the merge 4d image from all time points and then run the 3dFWHMx on that to get teh acf values ?
Could you please shed some light on the above points as well.
Best Wishes,
Neer
The
National Institute of Mental Health (NIMH) is part of the National Institutes of
Health (NIH), a component of the U.S. Department of Health and Human
Services.