Procedure would be something like the following:
1) Detect which columns have missing data
2) Construct 2-by-2 tables of missingness indicators for each combination. Depending on if any cell has a zero, create an indicator. Something like, non-monotone, monotonic (in what direction), always missing with the other
3) Fill in plot based on table results
It would be useful to have a plot measure (either through
matplotlibor through the console) that shows the missing data pattern (i.e. monotonic versus non-monotonic). Something resembling : https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5860553/figure/kwx350F1/Procedure would be something like the following: 1) Detect which columns have missing data 2) Construct 2-by-2 tables of missingness indicators for each combination. Depending on if any cell has a zero, create an indicator. Something like, non-monotone, monotonic (in what direction), always missing with the other 3) Fill in plot based on table results