Sorry but I am not able to answer regarding to your contrasts because I never conduct study with parametric stimulation.
I strength again my point about flexible factorial which is definitively a problem that you have to consider before to ask yourself if flexible factorial could manage incomplete repeated measures data).
It is statistically incorrect to include a group factor (population 1 vs. population 2) and a within-repeated factor (condition 1 vs. condition 2) in a flexible factorial in SPM. The reason come from the GLM estimation approach. The error term for group factor and within-repeated factor is consider in the same way which conduct to an imprecise modulizations. The factor group and the factor within should have two different error terms which is not handle by the flexible factorial of SPM. The manual “Contrast weights in flexible factorial design with multiple groups of subjects” of Jan Gläscher and Darren Gitelman is very handy and is an excellent reference to understand the manipulation of the flexible factorial. Except the example of “Two groups of subject” in page 7-10 which is bad due to the above reason.
In order to conduct this type of analysis, you have two ways:
1. Karl Friston's way: Subtract the two within conditions at the first level of each participant (condition 2 – condition 1) and make a 2nd level factorial design to compare the result of this subtraction contrast between groups.
2. If you want to make a mixed Design as your original plan, you should use a toolbox which handle this problem of error terms. I suggested you the SwE toolbox but there is others.
Hope it help even it is not directly what you asked.
Best regards
-----Original Message-----
From: João Simões <[log in to unmask]>
Sent: mercredi, 4 août 2021 18:59
To: [log in to unmask]; MOUTHON Michael <[log in to unmask]>
Subject: Re: Flexible Factorial ANOVA - incomplete repeated measures
Dear Michael,
I have already performed the analysis using the flexible factorial design (following the guidance of this paper: https://urldefense.com/v3/__https://www.researchgate.net/publication/267779738_Contrast_weights_in_flexible_factorial_design_with_multiple_groups_of_subjects__;!!Dc8iu7o!n2fR8NIxtyEDxww9qFrjjlOhID1GYziEHTPwFd_LGH2MfhPKVyc_oVO2VY7uy08Wk-RK0g$ ) only for the subjects with repeated measures. In the 1st level analysis, I created 4 regressors of interest: epoch for condition1, parametric modulator for condtion1, epoch for condition2 and parametric modulator for condtion2. Then, I created 2 contrasts, based on the parametric modulators of the two conditions. In the 2nd level, I used those contrasts in a flexible factorial design. Isn't that possible?
And if so, can it also accommodate the data with incomplete repeated measures?
Best regards
|