cannot run stats on a GLM that includes 2 tasks

[Remi-Gau]

at some point the pipeline will iterate over tasks and will try to run separate GLM for each tasks

[monique2208]

I just looking exactly for this possibility. This would be a really great feature as people occasionally put conditions across different runs and add this to the task label. I was wondering if it would be possible to specify task as list of tasks as well. I am only not sure what it would mean for the input columns, does it mean the event files of the different tasks need to share the column names and values, or would you add a prefix for the task?

[Remi-Gau]

quick reply: this is actually a bug I introduced recently

bidspm used to be able to combine several tasks into a single GLM

here is an example of model I had used: https://github.com/Remi-Gau/olfactory_fmri_blind/blob/master/src/analysis_func/models/model-default_smdl.json

you can see that several tasks are specified in the model

  "Input": {
    "task": [
      "olfid",
      "olfloc"
    ],
    "space": [
      "MNI152NLin2009cAsym"
    ]
  },

you can then just specify the events to include in the design matrix section as usual but in case you want to contrast things across tasks then at the moment the only way to do this would be by adding a prefix to the event name that specifies the tasks it came from

      "Model": {
        "X": [
          "trial_type.olfid_eucalyptus_left",
          "trial_type.olfid_eucalyptus_right",
          "trial_type.olfid_almond_left",
          "trial_type.olfid_almond_right",
          "trial_type.olfloc_eucalyptus_left",
          "trial_type.olfloc_eucalyptus_right",
          "trial_type.olfloc_almond_left",
          "trial_type.olfloc_almond_right",
          "trial_type.resp_03",
          "trial_type.resp_12",
          "trans_?",
          "rot_?",
          1
        ],
        "HRF": {
          "Variables": [
            "trial_type.olfid_eucalyptus_left",
            "trial_type.olfid_eucalyptus_right",
            "trial_type.olfid_almond_left",
            "trial_type.olfid_almond_right",
            "trial_type.olfloc_eucalyptus_left",
            "trial_type.olfloc_eucalyptus_right",
            "trial_type.olfloc_almond_left",
            "trial_type.olfloc_almond_right",
            "trial_type.resp_03",
            "trial_type.resp_12"
          ],
          "Model": "spm"
        },

I could try to adapt the demo from this mutlitask dataset: https://openneuro.org/datasets/ds000114/versions/1.0.2

but for many tasks the events were just called Task (who needs more informative event descriptors?...) so comparing across tasks will be difficult (but averaging across task should be easy).

[monique2208]

Great, I misunderstood but then this is exactly what I need, thank you! Averaging across task is all I need for now.

[Remi-Gau]

OK will try to fix soon and push a new release as well

[Remi-Gau]

closed by #1310