gatherStatistics: Gathering statistics from simulation()-Object

In mutoss: Unified Multiple Testing Procedures

Gathering statistics from simulation()-Object

Description

This function facilitates gathering statistics from the object returned by simulation().

Usage

gatherStatistics(simObject, listOfStatisticFunctions,
    listOfAvgFunctions)

Arguments

simObject	An object returned by simulation()
listOfStatisticFunctions	List of statistics that shall be calculated for the elements in simObject
listOfAvgFunctions	List of functions that will be used to summarize/average the calculated statistics for all elements in simObject$results with the same parameter constellation. If this is argument is missing no averaging will be done. Instead the resulting data.frame will keep one row for every object in simObject$results.

Details

For every simulation()-Object in simObject$results all statistics in listOfStatisticFunctions are calculated. If in addition listOfAvgFunctions is provided then the statistics of the objects that have the same parameter constellation are averaged by the functions in listOfAvgFunctions. The resulting data.frame will then keep only one row for every parameter constellation.

Value

statisticDF

A data.frame that can be of two different kinds. If listOfAvgFunctions is provided, then the resulting data.frame will have a row for every parameter constellation that occurs in the simObject$results. There will be columns for every parameter used. Also length(listOfAvgFunctions) * length(listOfStatisticFunctions) additional columns will be created. Every statistic is calculated for every applied procedure and then all values that belong to a specific parameter constellation are "averaged" by applying all function from listOfAvgFunctions. For example, suppose FDP is a function from listOfStatisticFunctions that calculates the realized false discovery proportion, that is number of true hypotheses that were rejected divided by the number of all rejected hypotheses. Also suppose mean and sd are functions in listOfAvgFunctions, then the mean and the standard deviation of the statistic FDP are calculated. If listOfAvgFunctions is not provided, then the resulting data.frame will have a row for every simObject$results. Every parameter will have its own column. Additional columns for every function in listOfStatisticFunction will be created.

Author(s)

MarselScheer

Examples

#' # this function generates pValues 
myGen <- function(n, n0) {
  list(procInput=list(pValues = c(runif(n-n0, 0, 0.01), runif(n0))), 
       groundTruth = c(rep(FALSE, times=n-n0), rep(TRUE, times=n0)))
}

# need some simulation()-Object I can work with 
sim <- simulation(replications = 10, list(funName="myGen", fun=myGen, n=200, n0=c(50,100)), 
list(list(funName="BH", fun=function(pValues, alpha) BH(pValues, alpha, silent=TRUE), 
  alpha=c(0.25, 0.5)),
list(funName="holm", fun=holm, alpha=c(0.25, 0.5),silent=TRUE)))

# Make my own statistic function
NumberOfType1Error <- function(data, result) sum(data$groundTruth * result$rejected)

# Get now for every object in sim$results one row with the number of Type 1 Errors
result.all <- gatherStatistics(sim, list(NumOfType1Err = NumberOfType1Error))

# Average over all sim$results-Objects with common parameters 
result1 <- gatherStatistics(sim, list(NumOfType1Err = NumberOfType1Error), list(MEAN = mean))
print(result1)  
result2 <- gatherStatistics(sim, list(NumOfType1Err = NumberOfType1Error), 
             list(q05 = function(x) quantile(x, probs=0.05), 
             MEAN = mean, q95 = function(x) quantile(x, probs=0.95)))
print(result2)

# create some plots
require(lattice)
histogram(~NumOfType1Err | method*alpha, data = result.all$statisticDF)
barchart(NumOfType1Err.MEAN ~ method | alpha, data = result2$statisticDF)

mutoss documentation built on March 31, 2023, 8:46 p.m.