Documentation/To Dos

[bakermoran]

[bakermoran]

• clean up/refactor loss function. Add new functionality
• on the cdf plot, dont despine the left (when i try this it
  • applies to all of them)
• add unit tests
• add ability to not sample, but run with seeded random numbers
  • (scipy or numpy)
• allow someone to input a lot of data, but select out two
  • variants to plot
    • ex: data contains var1, var2, var3, but want to only test var1 and var3
    • for some reason, this doesnt work:
      • self.raw_data =
        • raw_data.loc[raw_data[bucket_col_name].isin(variant_bucket_names)]
  • or else do this for now: make sure number of buckets in dataframe match
    • number of buckets in variant_bucket_names
• debug should actually output diagnostic info of the model
• in plot_positive_lift, if not self.compare_variants should not be a requirement
• naming of posteriors/priors is wrong. some are post some are prior.