Adds a new log file filter-/query-based selection of trials for constructing evokeds (instead of supply log file column names and creating evoked for all combinations of conditions)
That's useful for (a) incorporating continuous columns (e.g., RT > 200 ms) and (b) avoiding an explosion of the number of evoked in case of many combinations (some of which might not be of interest)
It required a re-factoring of tfr_subtract_evoked for subtracting the by-condition ERP before computing TFR. The new approach will simply subtract the grand average across all conditions if average_by is None or otherwise subtracts the relevant (filter-/query-based) evoked
The old API (i.e., supplying column names instead of a dictionary with query strings) continues to work from now but might get deprecated in the future
evokeds
(instead of supply log file column names and creating evoked for all combinations of conditions)tfr_subtract_evoked
for subtracting the by-condition ERP before computing TFR. The new approach will simply subtract the grand average across all conditions ifaverage_by
isNone
or otherwise subtracts the relevant (filter-/query-based) evokedFixes #107