Hi Michelle,

Please see below:

On 25 October 2016 at 16:50, Michelle VanTieghem <[log in to unmask]> wrote:

Hi Anderson,

Thanks for this model! To clarify the statistics underlying this model, is this performing a mixed effects model for the 2(between) x 3(within) interaction, or a repeated measures ANOVA?

These names don't really help much. The best is to show the designs. In any case, the first is testing only within-subject effects, including the interaction, which is also a within-subject effect, whereas the second is testing between-subject effects. It uses repeated measurements. It is an analysis of variance.

I have a couple of questions regarding the implementation of this model in PALM, since I have not used PALM before. The design.ods you provided actually contains 2 separate files, and each of those files includes the design.mat and design.con for (a) the 2x3 interaction and (b) the main effect of group across all 3 conditions.

Exactly.

(1) I cannot find an example in PALM user information where one can enter 2 separate design.mats and design.cons into the same command. Are they supposed to be entered as separate files, or is there a way to combine the files in a way that will perform both models simultaneously?

Use something as:

palm -i data.nii -d design1.mat -t design1.con -d design2.mat -t design2.con -eb blocks.csv -whole -within -corrcon [...]

(2) The EB column represents the exchangability blocks, within subjects. In the example provided, there should be a separate "blocks.csv" file input for the PALM command. [[palm -i modality1_4d.nii -d design.mat -t design.con -eb blocks.csv]] What is supposed to go into the blocks.csv file? If the EBs are already included within my design.mat file, do I also need a blocks.csv file?

Yes.

(3) I have not added F-tests to a design.con file before (only in the GUI). Is this already in the appropriate format such that the F-test is indicated by the final column in the design.con file, and the rows with no F-test indicated are just empty?

You can ignore the F-tests shown in the design. Use instead the option -corrcon. If however you were to create them, they are specified as .fts files in the same way as in randomise or FEAT.

(4) I would also like to include covariates of sex and Age (continuous). Can you advise how to modify this design to account for continuous covariates? Are they just added as separate EVs within the design.mat file?

If age and sex don't change between the 3 measurements per subject, these don't need to be entered in design1, as the subject-specific EVs already account for that. For design2, you can add extra EVs for age and sex.

(5) In addition, I would like to run a separate model to test the Age x Group x Emotion interaction. Is it possible to test a 3-way interaction with a continuous variable within this model framework in PALM?

Yes, it is. Please see design3 here: https://dl.dropboxusercontent.com/u/2785709/outbox/mailinglist/design_michelle.ods

Design1 and design2 are the same.

All the best,

Anderson