pval_aucpr: Area under the precision-recall curve

In OchoaLab/simtrait: Simulate Complex Traits from Genotypes

Area under the precision-recall curve

Description

Calculates the Precision-Recall (PR) Area Under the Curve (AUC) given a vector of p-values and the true classes (causal (alternative) vs non-causal (null)). This is a wrapper around PRROC::pr.curve(), which actually calculates the AUC (see that for details).

Usage

pval_aucpr(pvals, causal_indexes, curve = FALSE)

Arguments

pvals	The vector of association p-values to analyze. NA values are allowed in input, are internally set to 1 (worst score) prior to AUC calculation (to prevent methods to get good AUCs by setting more cases to NA). Non-NA values outside of [0,1] will trigger an error.
causal_indexes	The vector of causal indexes, defining the true classes used for AUC calculation. Values of causal_indexes as returned by sim_trait work. There must be at least one causal index and at least one non-causal case.
curve	If FALSE (default), only scalar AUC is returned. If TRUE, then curve = TRUE is passed to PRROC::pr.curve() and the full object (class PRROC) is returned (see below).

Value

If curve = FALSE, returns the PR AUC scalar value. If curve = TRUE, returns the PRROC object as returned by PRROC::pr.curve(), which can be plotted directly, and which contains the AUC under the named value auc.integral.

However, if the input pvals is NULL (taken for case of singular association test, which is rare but may happen), then the returned value is NA.

See Also

PRROC::pr.curve(), which is used internally by this function.

pval_power_calib() for calibrated power estimates.

Examples

# simulate truly null p-values, which should be uniform
pvals <- runif(10)
# for toy example, take the first two p-values to be truly causal
causal_indexes <- 1:2
# calculate desired measure
pval_aucpr( pvals, causal_indexes )

OchoaLab/simtrait documentation built on April 19, 2024, 7:36 p.m.