CohortDiagnostics
CohortDiagnostics copied to clipboard
Compare Chararacterization: StDiff enhancements
Two items on standardized difference:
-
All StDiff values should be listed with 2 decimals. currently we only show 1 decimal point:
-
If target or comparator has 0%, but the other cohort has a value >0%, we should be able to compute StDiff. Currently, StDiff is NULL in these cases: (need to coalesce the null cohort to have mean = 0, stdev = 0, then the standard StDiff formula will be fine).
Note, I think this may be specifically influenced by the 'Min Covariate Mean' user input parameter. It could be there is non-zero values that just fall below the threshold.
The calculation has been fixed as per screenshot below:
however, I'm keeping this issue open as currently the operation is slow as legacy code pulls down data the target and comparator separately - Instead it would be significantly more performant to pull it down for both doing the standard diff calculation in SQL.