Ratio With Shrinkage
Last updated
Last updated
Performs empirical bayesian estimation with shrinkage towards a beta prior. Given a dataset with a numerator and a denominator, will calculate the raw ratio as numerator / denom, as well as provide an adjusted ratio that shrinks the ratio towards the observed beta prior.
This is a simplified version that establishes the priors directly from the data given a min_cutoff count of observations.
NOTE: your data should already be aggregated before performing this operation.
Name | Type | Description | Is Optional |
---|---|---|---|
numerator
column
A column that is pre-aggregated to contain the count of positive cases
denom
column
A column that is pre-aggregated to contain the count of ALL cases
min_cutoff
int
Enter a minimum value to limit the denominator when creating the prior estimates. Example: if estimating a batter's hitting percentage, entering 500 would limit the estimation of the priors to be only for batters with over 500 at-bats.