Iss20

[Thiyaghessan]

I think I did all the things!

1. Calculate the mean absolute difference in proportions for each combination of classes for 1-, 2-, and 3-way marginals.
2. Highlight specific margins and cells where the utility is poor.
3. Allow for different structures (longitudinal data, hierarchical data).
4. Allows for sampling from a subset of margins when the combinatorics create too many potential 2- and 3-way marginals.
5. Allows for prioritizing variables if picking a subset of the 2- and 3-way marginals (e.g., include all marginals that use age).
6. Support sample weights.
7. Allow for discretizing continuous variables using fixed binwidths, ntiles, and univariate clustering.