superbData: Obtain summary statistics with correct error bars.

In dcousin3/superb: Summary Plots with Adjusted Error Bars

Obtain summary statistics with correct error bars.

Description

The function suberbData() computes standard error or confidence interval for various descriptive statistics under various designs, sampling schemes, population size and purposes, according to the suberb framework. See \insertCitecgh21superb for more.

Usage

superbData(
  data,
  BSFactors = NULL,
  WSFactors = NULL,
  WSDesign = "fullfactorial",
  factorOrder = NULL,
  variables,
  statistic = "mean",
  errorbar = "CI",
  gamma = 0.95,
  adjustments = list(purpose = "single", popSize = Inf, decorrelation = "none",
    samplingDesign = "SRS"),
  preprocessfct = NULL,
  postprocessfct = NULL,
  clusterColumn = ""
)

Arguments

data	Dataframe in wide format
BSFactors	The name of the columns containing the between-subject factor(s)
WSFactors	The name of the within-subject factor(s)
WSDesign	the within-subject design if not a full factorial design (default "fullfactorial")
factorOrder	Order of factors as shown in the graph (x axis, groups, horizontal panels, vertical panels)
variables	The dependent variable(s)
statistic	The summary statistic function to use
errorbar	The function that computes the error bar. Should be "CI" or "SE" or any function name. Defaults to "CI"
gamma	The coverage factor; necessary when errorbar == "CI". Default is 0.95.
adjustments	List of adjustments as described below. Default is adjustments = list(purpose = "single", popSize = Inf, decorrelation = "none", samplingDesign = "SRS")
preprocessfct	is a transform (or vector of) to be performed first on data matrix of each group
postprocessfct	is a transform (or vector of)
clusterColumn	used in conjunction with samplingDesign = "CRS", indicates which column contains the cluster membership

Details

The possible adjustements are the following

• popsize: Size of the population under study. Defaults to Inf

• purpose: The purpose of the comparisons. Defaults to "single". Can be "single", "difference", or "tryon".

• decorrelation: Decorrelation method for repeated measure designs. Chooses among the methods "CM", "LM", "CA" or "none". Defaults to "none".

• samplingDesign: Sampling method to obtain the sample. implemented sampling is "SRS" (Simple Randomize Sampling) and "CRS" (Cluster-Randomized Sampling).

Value

a list with (1) the summary statistics in summaryStatistics (2) the raw data in long format in rawData (using numeric levels for repeated-measure variables).

References

\insertAllCited

Examples

# Basic example using a built-in dataframe as data; 
# by default, the mean is computed and the error bar are 95% confidence intervals
# (it also produces a $rawData dataframe, not shown here)
res <- superbData(ToothGrowth, BSFactors = c("dose", "supp"), 
  variables = "len") 
res$summaryStatistics

# Example introducing adjustments for pairwise comparisons 
# and assuming that the whole population is limited to 200 persons
res <- superbData(ToothGrowth, BSFactors = c("dose", "supp"), 
  variables = "len",  
  statistic = "median", errorbar = "CI", gamma = .80,
  adjustments = list( purpose = "difference", popSize = 200) )
res$summaryStatistics

# Note that you can achieve the same with formulas
superb( len ~ dose + supp, ToothGrowth, showPlot=FALSE)

dcousin3/superb documentation built on Dec. 23, 2024, 8:58 p.m.