calculate_rmse_rse: Calculate RMSE and RSE for predictive estimates

In njhenry/covidemr: Tools for Mapping COVID Excess Mortality

Description

Given a dataset, calculate Root Mean Squared Error and Relative Squared Error (both weighted and unweighted) for predictions across a dataset. The observed data should come in the form of separate numerator and denominator columns, while the estimates should be a single field estimating a rate (i.e. already normalized by denominator)

Usage

calculate_rmse_rse(
  in_data,
  num_field,
  denom_field,
  est_field,
  group_fields = NULL
)

Arguments

in_data	Input data.table
num_field	Numerator field for the observed data
denom_field	Denominator field for the observed data
est_field	Estimator field, presented as a rate (num/denom)
group_fields	[optional, default NULL] If the predictive validity metrics should be grouped, list the fields to group them by here. If NULL (the default), the predictive validity metrics will be calculated across the entire dataset

Value

A data.table with the following fields: - 'rmse': Root mean squared error, unweighted - 'rmse_weighted': Root mean squared error, weighted by population size - 'rse': Relative squared error, unweighted - 'rse_weighted': Relative squared error, weighted by population size - Any grouping columns specified in the function arguments

njhenry/covidemr documentation built on Feb. 2, 2022, 2:31 a.m.