Skip to contents

Computes concordance indices for model evaluation

Source: R/concordance.R
concordance_indices.Rd

These functions compute different concordance indices for use/availability data in a train-validate-test model evaluation context.

Usage

conditionalBoyce(
  x,
  method = c("pearson", "kendall", "spearman")[1],
  plotit = FALSE,
  errors = TRUE,
  warnings = TRUE
)

conditionalSomersD(x, errors = TRUE, warnings = TRUE)

conditionalAUC(x, errors = TRUE, warnings = TRUE)

AUC(x, errors = TRUE, warnings = TRUE)

coxnet.deviance(x, errors = TRUE, warnings = TRUE)

Cindex(x, errors = TRUE, warnings = TRUE)

Arguments

x

[data.frame]
A data frame with three columns: x (predicted values), y (use/available response, 1/0), and strat (stratum identifier).

method

[character(1)="pearson"]{"pearson","kendall","spearman"}
Correlation method used to compute the Boyce index. Only used in conditionalBoyce().

plotit

[logical(1)=FALSE]
Whether to plot the bin frequency distribution. Only used in conditionalBoyce().

errors

[logical(1)=TRUE]
Whether to raise an error if the number of non-empty bins is insufficient for a reliable estimate.

warnings

[logical(1)=TRUE]
Whether to emit warnings for unequal strata lengths or low bin counts.

Value

A single numeric value representing the concordance metric:

  • conditionalBoyce(): Pearson/Kendall/Spearman correlation of bin frequencies.

  • conditionalSomersD(): Somers' D statistic (range -1 to 1).

  • conditionalAUC(): AUC derived from Somers' D (range 0 to 1).

  • AUC(): AUC from pROC::auc(), ignoring strata.

  • coxnet.deviance(): Cox partial deviance via glmnet::coxnet.deviance().

  • Cindex(): Concordance index via glmnet::Cindex().

Details

The function conditionalAUC() is the implementation of the AUC as related to the Somers' D index. It accounts for strata, ideal for conditional logistic regression, but is under testing. AUC() uses pROC::auc() and does not account for strata. coxnet.deviance and Cindex are wrappers for glmnet::coxnet.deviance() and glmnet::Cindex() using the same argument structure as the other concordance functions.