Bidspm with 3 groups

[FioritoAnna]

Hi,

I am working on a neuroimaging dataset with three groups (control, patients, relatives) and I am using bidspm.
I am aware that bidspm only work for one-sample or two-sample t-test, and I would like to perform 2-by-2 comparisons (controls vs patients, patients vs relatives, relatives vs controls).

For instance to look at controls >patients in my bids stat model for between group comparisons at dataset level I would use something like:

{
      "Level": "Dataset",
      "Name": "between_group",
      "Description": "two sample t-test",
      "GroupBy": [
        "contrast"
      ],
      "Model": {
        "Type": "glm",
        "X": [
          1,
          "Group"
        ]
      },
      "DummyContrasts": {
       "Test": "t"
     },
      "Contrasts": [
        {
          "Name": "hc_gt_sz",
          "ConditionList": [
             "Group.controls",
             "Group.patients",
             "Group.relatives"             
          ],
          "Weights": [
            1,
            -1,
            0
          ],
          "Test": "t"
        }
      ]
    }

However, this is not working (I can only make it work if I delete individuals subjects folders of relatives in my bids, bidspm-preproc and bidspm-stats directories) and use ;

 "ConditionList": [
             "Group.controls",
             "Group.patients"
]

Is there a more straightforward way of using bidspm with 3 groups?

Thank you!
All the best,
Anna

[github-actions]

Thank you for your issue. Give us a little time to review it.

PS. You might want to check the FAQ if you haven't done so already.

This is an automated reply, generated by FAQtory

[Remi-Gau]

I can only make it work if I delete individuals subjects folders of relatives in my bids, bidspm-preproc and bidspm-stats directories

Ouch !!! that's a bit drastic.
Sorry about that.

Will try to set a minimal example to see if I can find a workaround.

I would like to perform 2-by-2 comparisons (controls vs patients, patients vs relatives, relatives vs controls)

Not sure I follow, that seems more like a one-way (1 by 3) anova with some planned contrats. You only have a "group" factor with 3 levels, right?

[FioritoAnna]

That would be nice, thanks!

Since my understanding is that it's not possible to perform ANOVA in bidspm, I thought that instead of first performing a one-way ANOVA and then post-hoc t-tests I can directly test my hypothesis with t-tests (with 2 groups I would only have to perform one t-test - patients vs controls- while with 3 groups I will perform 3 separate t-tests - controls vs patients, patients vs relatives, relatives vs controls but to do so I need to be able to "ignore" one of the three groups when it is not employed in a given comparison, which I can't do for now).

Not sure this answer to your question but I have a between-factor (3 groups) and a task-related within-factor (my task have three 2-level factors which are attention high/attention low, valence positive/valence negative, congruent/incongruent).

Don't hesitate to tell me if this is not clear

[FioritoAnna]

Hi,
Any follow-up on this issue?
Thanks,
Anna

[Remi-Gau]

@FioritoAnna

sorry I never got back to you on this.

I am finally implementing this in #1296 by adding the possibility to create one-way anova at the group level to compare different groups.

Took me quite a while and a lot of changes but now bidspm should automatically decide between 2 samples T Test or one way anova depending on the number of groups it finds in your participants.tsv