plotAccuracy: Plot predictive accuracy
In DudbridgeLab/AVENGEME: Analysis of polygenic scoring methods

Description Usage Arguments Details Value Author(s) References Examples

View source: R/plotAccuracy.R

Plots the AUC or the R2 as a function of training sample size.

plotAccuracy(xlim = c(1, 500), ylim = 0, nsnp = 1e+05, vg1 = 0,
  pi0 = 0, cov12 = NA, fix = TRUE, binary = FALSE, prevalence = 0,
  sampling = 0.5, r2gx = 0, corgx = 0, r2xy = 0,
  adjustedEffects = FALSE, plot = TRUE, col = "black", breakeven = 0.5,
  lty = 1)

`xlim`	Vector of 2 elements, giving the range of sample size to display on the x-axis, in 1000s. For binary traits this is the number of cases.
`ylim`	Range of AUC/R2 to display on y-axis.
`nsnp`	Number of independent SNPs in the gene score.
`vg1`	Proportion of variance explained by genetic effects in the training sample.
`pi0`	Proportion of markers with no effect on the training trait.
`cov12`	Covariance between genetic effect sizes in the two samples. If the effects are fully correlated then cov12<=sqrt(vg1). If the effects are identical then cov12=vg1 (default).
`fix`	TRUE if the same genetic model is assumed for the training and target samples.
`binary`	TRUE if the training trait is binary. By default, the target trait is binary if the training trait is; otherwise binary should be a vector with two elements for the training and target samples respectively.
`prevalence`	For a binary trait, prevalence in the training sample. By default, prevalence is the same in the target sample. Otherwise, prevalence should be a vector with two elements for the training and target samples respectively.
`sampling`	For a binary trait, case/control sampling fraction in the training sample. By default, sampling equals the prevalence, as in a cohort study. If the sampling fraction is different in the target sample, sampling should be a vector with two elements for the training and target samples respectively.
`r2gx`	Proportion of variance in environmental risk score explained by genetic effects in training sample.
`corgx`	Genetic correlation between environmental risk score and training trait.
`r2xy`	Proportion of variance in training trait explained by environmental risk score.
`adjustedEffects`	TRUE if polygenic and envrionmental scores are combined as a weighted sum. If FALSE, the scores are combined as an unweighted sum even if they are correlated.
`plot`	TRUE is a new plot is to be drawn, otherwise draw lines on the existing plot.
`col`	Colour in which to plot.
`breakeven`	Value of AUC/R2 for which the minimum sample size will be estimated.
`lty`	Line type parameter for R plots.

AUC is plotted for binary traits, R2 for quantitative traits. At each point, the p-value threshold is identified for selecting markers into the polygenic score, such that the AUC or R2 is maximised.

A list with the following elements:

limit Value of AUC/R2 at the maximum sample size plotted.
breakeven Sample size at which the AUC/R2 exceeds the value specified by the breakeven parameter.
plimit Optimal P-value threshold at the maximum sample size plotted.

Frank Dudbridge

Dudbridge F (2013) Power and predictive accuracy of polygenic risk scores. PLoS Genet 9:e1003348