Seeking help/clarification re: 3dMVM, 3dLME or a more appropriate alternative to my data

AFNI Message Board
1

Hi Gang (or anyone available to help me).

First, I hope this finds you well during these crazy times. Second, I am writing to you hoping that you can help me determine whether 3dMVM or 3dLME (or something else) is suited for my data. I have the following mixed factorial repeated measures event-related fMRI design:

  • 5 functional runs
  • 180 trials per subject
  • within person variable with 2 levels: emotion
  • within person variable with 2 levels: situation
  • between person variable with 2 levels: nationality
  • quantitative variable: time spent in the United States (a variable of interest)
  • quantitative variable: age (a variable of NO interest that we’d like to control for since it is correlated with the above variable, time spent in the US)

I am interested in running whole-brain analyses that test for significant main and interaction effects for the variables of interest. Specifically, I’d like to test the following and their ensuing post-hoc comparisons:

whole-brain BOLD ~ emotionsituationnationality
whole-brain BOLD ~ emotionsituation[time spent in the US] + age

Importantly, if the interaction effects prove non-significant, is it safe to still interpret the main effects using 3dMVM or 3dLME? Or would I need to rerun the analyses and omit the interaction terms?

Please let me know if there’s anything I can clarify.
Thank you so much.
Joseph

2

Joseph,

Clarifications:

  • Are the number of runs and the number of trials are variables/factors at the population level?
  • Does age a single or multiple values per subject?
3

  • The number of runs and the number of trials are technically at the individual level (each trial, for instance, was a combination of the emotion factor and situation factor; there are 180 trials per subject, 36 trials per run). I’m not sure if this matters, but preprocessing and first-level analyses were done in SPM.

  • Subjects have single age values, so for instance, subject #1 is 30 years old. There are 50 subjects, so 50 age values.

Let me know if that doesn’t clarify your questions. Thanks!

4

Gang, I may have misunderstood your first question. The number of runs and the number of trials make up the two fixed, within subject factors: emotion (2 levels) and situation (2 levels). So ultimately, yes, the number of runs and the number of trials are variables/factors at the population level. See my previous points, however, and let me know what you think.

Thank you again.

5

Joseph, I assume that runs and trials are dealt with at the subject level. In other words, the variables at the population level are the three within-subject factors plus a between-subject quantitative variable. Use 3dMVM to handle such a straightforward ANCOVA structure even though 3dLME/3dLMEr would work too.