Reevaluate hitcontinner proportional weights

[madison-feshuk]

Based on the #88 and related analysis reviewing this issue and fix (Rmd and html). ToxCast team approved this fix for its invitrodb v4.2 release, but had questions about current implementation of hitc as the product of 3 proportional weights. It is evident by its non-normal distribution that hitc can be easily skewed to either 0 or 1.

Are there additional proportional weights that should be added to the continuous hitcall calculations? Bayesian approach?

• Consider noise
• Consider relative efficacy (i.e. scaled top) within endpoint
• Consider degree of equivocality (relativity of top_over_cutoff i.e. for borderline case)