as part of creating the design matrix with uf_designmat, unfold checks the rank of the design matrix and gives out a warning if the rank is smaller than the number of columns of the DM.
check_rank(EEG.unfold.X)
However, this seems to happen even with uncorrelated predictors. The reason appears to be that when creating the (original, non-time-expanded) DM, all EEG.event entries in the original EEG dataset are represented as one row in the DM, regardless of whether the event type is part of the model formula or not. These zero-only rows then appear to affect the rank computation.
Possible fixes:
- ? remove zero-only rows from non-modelled event types from DM (may cause downstream problems...)
- ? exclude zero-only rows from rank check
as part of creating the design matrix with
uf_designmat, unfold checks the rank of the design matrix and gives out a warning if the rank is smaller than the number of columns of the DM.check_rank(EEG.unfold.X)However, this seems to happen even with uncorrelated predictors. The reason appears to be that when creating the (original, non-time-expanded) DM, all
EEG.evententries in the original EEG dataset are represented as one row in the DM, regardless of whether the event type is part of the model formula or not. These zero-only rows then appear to affect the rank computation.Possible fixes: