Clarify raw_haemo in group-level GLM tutorial

[mattrussell2]

In the group-level GLM tutorial, raw_haemo is returned after running each individual participant separately.

df_roi = pd.DataFrame()  # To store region of interest results
df_cha = pd.DataFrame()  # To store channel level results
df_con = pd.DataFrame()  # To store channel level contrast results

for sub in subjects:  # Loop from first to fifth subject
    # Create path to file based on experiment info
    bids_path = dataset.update(subject=sub)

    # Analyse data and return both ROI and channel results
    raw_haemo, roi, channel, con = individual_analysis(bids_path, sub)

    # Append individual results to all participants
    df_roi = pd.concat([df_roi, roi], ignore_index=True)
    df_cha = pd.concat([df_cha, channel], ignore_index=True)
    df_con = pd.concat([df_con, con], ignore_index=True)

This is not stored anywhere, nor aggregated across participants. Yet, later in the tutorial, raw_haemo is used in group topographic visualizations and contrast results. As the code is written, I believe that the last participant's raw data will be used, rather than some form of group average. Is this expected behavior? Thanks.

[mattrussell2]

After thinking about this a bit more, I think I was mistaken in that the raw_haemo object here is used primarily to extract the channel locations, which should be consistent across participants. If this is correct, I'd be happy to make a clarifying comment to this effect in the tutorial if that would be helpful.

[drammock]

I think wherever you see raw_haemo.copy().pick(picks="hbo") (or picks="hbr") that is only being used for e.g. channel location info, and the plotted data values are coming from ch_model_df, con_model_df, etc. Is that what you eventually concluded? I think a good way to clarify this would be to create 2 new variables, maybe:

raw_hbr_chans = raw_haemo.copy().pick(picks="hbr")
raw_hbo_chans = raw_haemo.copy().pick(picks="hbo")

and just re-use those variables throughout the rest of the tutorial (and add to that snippet whatever explanatory comment you think it needs). Assuming I'm right that those objs aren't getting mutated by the plotting functions (they shouldn't be), a nice side effect is that it will eliminate a lot of unnecessary copies.