fwerPowerFdrPower: Simulate FWER, POWER, FDR, and POWER

In mshasan/OPWpaper: Reproduce figures and simulations of OPWeight

Description

This function simulate the Family Wise Error Rate (FWER) and the corresponding Power, and the False Discovery Rate (FDR) and the corresponding Power for the different effect sizes

Usage

fwerPowerFdrPower(i, simu, null, corr = 0, cv = 0, alpha = 0.05,
  groupSize = 100, effectType = c("continuous", "binary"),
  covariateEffectVec, datWeightByNull)

Arguments

i	Integer, i-th effect size of a vector of effects
simu	Integer, number of replications
null	Numeric, proportion of the true null hypothesis
corr	Numeric, correlation between the test statistics
cv	Numeric, coefficient of variation of the test statistics
alpha	Numeric value of the significance threshold
groupSize	Integer, number of test statistics per group
effectType	Character ("continuous" or "binary"), type of effect sizes
covariateEffectVec	A numeric vector of different covariate-effect size
datWeightByNull	A numeric matrix of weights, each column corresponds to an covariate-effect size

Details

This function simulate Family Wise Error Rate (FWER) and corresponding Power, and False Discovery Rate (FDR) and the corresponding Power for the different effect sizes

Value

A matrix of 16 rows containing information about FWER, POWER, FDR, and POWER (4 rows for each item)

Author(s)

Mohamad S. Hasan, [e-mail]

References

Hasan and Schliekelman (2017)

See Also

weight_byEffect_cont ranksProb_byEffect

Examples

# vector of covariate-effect sizes
covariateEffectVec <- c(1, 1.5, 2)

# compute probability matrix
ranksProb_byEffect <- sapply(1:length(covariateEffectVec), ranksProb_byEffect,
             null = .9, m = 100, covariateEffectVec = covariateEffectVec)

# compute weights
weightByEffect <- sapply(1:length(covariateEffectVec), weight_byEffect_cont,
                   alpha = .05, null = .9, m = 100, delInterval = .01,
                   covariateEffectVec = covariateEffectVec,
                   datByNull = ranksProb_byEffect)

simuVal = 3  # in the actual case use at least simVal = 1000
result <- sapply(1:length(covariateEffectVec), fwerPowerFdrPower, simu = simuVal,
             null = .5, corr = 0, cv = 0, alpha = .05, groupSize = 100,
             effectType = "continuous", covariateEffectVec = covariateEffectVec,
             datWeightByNull = weightByEffect)

mshasan/OPWpaper documentation built on March 3, 2021, 7:02 a.m.