UNCDEPENdLab / fmri.pipeline

R package for running full fMRI GLM analyses in FSL and other software.
GNU General Public License v3.0
4 stars 0 forks source link

Make EV_block Contrast Names Consistent with block Contrast Names #4

Open wasciutto opened 1 year ago

wasciutto commented 1 year ago

Naming of "EV block" contrasts not consistent "block" contrasts

See C1-C3 vs C5-C7 here: image

Looks like EV_block.trial_typeBigBoost should be EV_block.trial_type.BigBoost here:

wasciutto commented 1 year ago

My config if needed:

# define columns that can be selected in different parts of the model specification
onsets:
  - block_onset

durations:
  - block_duration

wi_factors:
  - trial_type

# events must consist of name, onset, and duration
events:
  block:
    onset: block_onset
    duration: block_duration

# if not otherwise specified, signals will default to unit height regressors with no HRF normalization
signals:
  block:
    event: block
    normalization: none
    value_fixed: 1
    wi_factors: trial_type
    wi_formula: ~ trial_type - 1

l1_models:
  basic_block:
    signals:
      - block
    contrasts:
      include_diagonal: no
michaelhallquist commented 1 year ago

Hi Will,

For better or worse, R's convention in linear models with factors is to omit a '.' in the coefficients for the dummy codes. Given this, I'm likely to keep the naming convention the same (no '.').

wasciutto commented 1 year ago

Gotchya, thanks for confirming that. Removing the "bug" tag because it's intentional, but I still feel it's worth considering how to make the contrast names consistent with each other at least on the presentation side, so I'll leave it open incase we want to re-vist in future (or until I learn the codebase enough to realize this makes no sense 🙃 )