calculate_epp: Calculate EPP Meta-Scores for All Players

In ModelOriented/EloML: Elo rating system for machine learning models

Calculate EPP Meta-Scores for All Players

Description

Calculate EPP Meta-Scores for All Players

Usage

calculate_epp(
  results,
  decreasing_metric = TRUE,
  compare_in_round = TRUE,
  keep_columns = FALSE,
  keep_model = FALSE,
  reference = NULL,
  keep_data = TRUE,
  estimation = "glmnet"
)

Arguments

results	Data frame. Results for one Round Data should be in the following format. First 3 columns should correspond to: Player, Round, Score. See more in 'details' section.
decreasing_metric	Logical. If TRUE used Score is considered as decreasing, that means a Player with higher Score value is considered as winner. If FALSE used Score will be considered as increasing.
compare_in_round	Logical. If TRUE compares Players only in the same fold. If FALSE compares Players across folds.
keep_columns	Logical. If TRUE original data frame with new 'epp' column will be returned.
keep_model	Logical. If TRUE logistic regression model to compute EPP will be returned.
reference	Character. Name of the Player that should be a reference level for EPP Meta-scores. It should be a name of one of the Players from 'results' data frame. If NULL, none of the Players will be chosen.
keep_data	If all the meta-data should be kept in result.
estimation	Method of estimating EPP coefficients, 'glm' or 'glmnet'.

Details

Format of the data frame passed via results parameter. First column should correspond to a Player. Second column corresponds to indexes of Rounds As EPP is based on Elo rating system, power of Player is assessed by comparing its results with other Players on multiple Rounds. Therefore, each Player should be evaluated on multiple Rounds. Indexes of Rounds should be in this column. And should match across Players. Third column contains Score used to evaluate players. It can be both decreasing or increasing metric. just remember to set the decreasing_metric parameter accordingly. The following columns can be of any kind.

Naming convention, such as Player, Rounds, etc. comes from Gosiewska et al. (2020).

Value

epp_results

Examples

library(EloML)
data(auc_scores)
calculate_epp(auc_scores[1:400,])

ModelOriented/EloML documentation built on June 22, 2022, 4:53 a.m.