ffmanovatest {ffmanova}R Documentation

50-50 MANOVA testing

Description

The function performs 50-50 MANOVA testing based on a matrix of hypothesis observations and a matrix of error observations.

Usage

ffmanovatest(modelData, errorData, stand = 0, part = c(0.9, 0.5),
             partBufDim = 0.5, minBufDim = 0, maxBufDim = 1e+08,
             minErrDf = 3, cp = -1)

Arguments

modelData matrix of hypothesis observations
errorData matrix of error observations
stand Standardisation (0 or 1) of responses
part The variance explained required when choosing the number of components for testing. The default value is 0.5, but to choose a single component 0.9 is required.
partBufDim tuning parameter for the number of buffer components
minBufDim minimum (if possible) number of buffer components
maxBufDim maximum number of buffer components
minErrDf minimum number of "free dimensions"
cp correction parameter when "few" responses

Details

modelData and errorObs correspond to hypObs and errorObs calculated by xy_Obj.

Value

A list with components

exVar1 variance explained by dimY components
exVar2 variance explained by dimY+bufferDim components
dim dimension of final "MANOVA-space"
dimX the ordinary degrees of freedom for the test
dimY number of components for testing
bufferDim number of buffer components
D test statistic: Wilks' Lambda
E test statistic: Roy's Largest Root
A test statistic: Hotelling-Lawley Trace Statistic
M test statistic: Pillay-Bartlett Trace Statistic
pD p-value: Wilks' Lambda
pE p-value: LOWER BOUND for Roy's Largest Root
pA p-value: Hotelling-Lawley Trace Statistic
pM p-value: Pillay-Bartlett Trace Statistic

Author(s)

Øyvind Langsrud and Bjørn-Helge Mevik

See Also

ffmanova


[Package ffmanova version 0.1-1.1 Index]