0todd0000
commented
1 month ago We performed post-hoc analysis with the Bonferonni correction (like in ex1d_anova1_posthoc) but nothing came out significant.
One possible explanation is non-normality. If the data are non-normal, several strange results can appear. Please consider trying spm1d's nonparametric procedure (spm1d.stats.nonparam.anova2onerm).
Another possibility is that the post hoc procedure must match the factor type. If you are conducting pos hoc analysis on a within-subjects factor then you must use a repeated-measures (i.e., within-subjects) procedure for post hoc analysis, like spm1d.stats.ttest_paired. Otherwise between-subject post hoc analysis should be used, like spm1d.stats.ttest2.
Should we only consider the interaction (GM FL) and the time effect (Torque) without providing more details or can we, in some way, run further analysis ?
It's difficult to consider this issue without additional information. Please first try nonparametric analysis and/or checking the between- vs. within-subject nature of the post hoc analysis as described above. If that does not resolve this issue please repost.
Dear Todd,
First of all, thank you for providing SPM and for your availability to answer all our questions.
I have a series of question concerning anova2onerm and post hoc analysis. We recorded ankle passive torque and gastrocnemius medialis fascicle length (GM FL) across the ankle range of motion in 2 groups before and after 2 types of training. One group was composed of 14 subjects and the other of 12.
We used the anova2onerm function for Group (2) X Time (2) interection in these 2 variables.
For ankle torque, spm highlightsno Group X Time interaction but only a time effect on the first 45% of the ROM. For the GM fascicle length, we have a Group X Time interaction on the first 30% of the ROM and that's all.
And to be honest after that, I get a bit confused. We performed post-hoc analysis with the Bonferonni correction (like in ex1d_anova1_posthoc) but nothing came out significant.
Should we only consider the interaction (GM FL) and the time effect (Torque) without providing more details or can we, in some way, run further analysis ?
For exemple, when doing simple paired t-test between G1-PRE vs G1-POST on GM FL, we found a significant increased GM FL on the first 25% of the ROM but nothing between G2-PRE and G2-POST. For the torque, we found a decreas in both groups from PRE to POST on the first 20% (G1) and 25% (G2). I imagine that this is too simple to be accurate.
Thank you in advance, Best regards, Baptiste