PhenTestResult-class: Class "PhenTestResult"
In PhenStat: Statistical analysis of phenotypic data

Description Explore PhenTestResult object Slots Methods Author(s) Examples

A list-based S4 class for storing of model fitting results and other information like dependent variable, method used to build a model, etc. Helps to support operations needed for model fitting process and modelling/testing results storage. In the package PhenTestResult object is created by function testDataset.

PhenTestResults object instance contains the following slots:

1. "analysedDataset" contains subdataset that was used for the modelling/testing:

analysedDataset(phenTestResult)

2. "depVariable" contains dependent variable that was tested: getVariable(phenTestResult)

3. "refGenotype" contains reference genotype which is usually is wildtype: refGenotype(phenTestResult)

4. "testGenotype" contains test genotype: testGenotype(phenTestResult)

5. "method" contains method name that was used for modelling/testing: method(phenTestResult)

6. "transformationRequired" contains TRUE/FALSE value indicating transofrmation of dependent variable:

transformationText(phenTestResult) or phenTestResult@transformationRequired

7. "lambdaValue" contains lambda value for the Box-Cox transformation:

transformationText(phenTestResult) or phenTestResult$lambdaValue

8. "scaleShift" contains the value of scale shift for Box-Cox transformation:

transformationText(phenTestResult) or phenTestResult@scaleShift

9. "transformationCode" contains the code of transformation. Possible values:

0 - transformation is not applicable (methods "FE", "RR", "LR") or switched off by user 1 - transformation is not needed (1 is within the 95 2 - log transformation 3 - power transformation 4 - transformation is not performed since optimal lambda value is not found (-5 > lambda > 5)

transformationText(phenTestResult) or phenTestResult@transformationCode

10. "parameters" contains the parameters used during method application, e.g. pThreshold for MM and TF:

parameters(phenTestResult)

11. Modelling/testing results are stored in the sections "analysisResults": analysisResults(phenTestResult)

The contents of the analysisResults slot depend on framework that was used.

The results of MM and TF frameworks: 1. Equation used during linear modeling: analysisResults(phenTestResult)$equation

2. Batch effect significance: analysisResults(phenTestResult)$model.effect.batch

3. Variance equality: analysisResults(phenTestResult)$model.effect.variance

4. Weight effect significance: analysisResults(phenTestResult)$model.effect.weight

5. Sex effect significance: analysisResults(phenTestResult)$model.effect.interaction

6. Evidence of sex effect (p-value): analysisResults(phenTestResult)$model.output.interaction

7. Evidence of genotype effect (p-value): analysisResults(phenTestResult)$model.output.genotype.nulltest.pVal

8. Formula for the final genotype model: analysisResults(phenTestResult)$model.formula.genotype

9. Formula for the final null model: analysisResults(phenTestResult)$model.formula.null

10. Model fitting output: analysisResults(phenTestResult)$model.output

11. Model fitting summary: summary(analysisResults(phenTestResult)$model.output)$tTable

The results of FE and RR frameworks are stored as a list of htestPhenStat S4 objects: analysisResults(phenTestResult)[[1]].

Each one htestPhenStat object contains:

1. Output of Fisher Exact Test: pvalue(analysisResults(phenTestResult)[[1]]).

2. Effect size: ES(analysisResults(phenTestResult)[[1]]).

3. Name of the table analysed (all, males, females): analysedSubset(analysisResults(phenTestResult)[[1]]).

4. Comparison, used for RR only (High vs Normal/Low, Low vs High/Normal): comparison(analysisResults(phenTestResult)[[1]]).

5. Matrix of counts: matrixCount(analysisResults(phenTestResult)[[1]]).

analysedDataset:: Object of class "data.frame" ~~ analysed dataset
transformationRequired:: Object of class "logical" ~~ flag indicating was or not transformation performed
lambdaValue:: Object of class "numeric" ~~ Box-Cox transform lambda value
scaleShift:: Object of class "numeric" ~~ Box-Cox transform scale shift
transformationCode:: Object of class "numeric" ~~ code explaining the transformation output
depVariable:: Object of class "character" ~~ dependent variable
refGenotype:: Object of class "character" ~~ reference genotype
testGenotype:: Object of class "character" ~~ test genotype
method:: Object of class "character" ~~ analysis method
parameters:: Object of class "matrix" ~~ parameters of method
analysisResults:: Object of class "list" ~~ results of the analysis

getVariable: (accessor): Returns dependent variable
refGenotype: (accessor): Returns reference genotype
testGenotype: (accessor): Returns test genotype
method: (accessor): Returns method used
methodText: (accessor): Returns full name of the method used
parameters: (accessor): Returns parameteres used during the method application
analysisResults: (accessor): Returns analysis results depending on method
analysedDataset: (accessor): Returns analysed dataset
transformationText: (accessor): Returns the info about transformation
batchIn: Returns TRUE if the batch is in the dataset, FALSE otherwise
weightIn: Returns TRUE if the weight is in the dataset, FALSE otherwise
noSexes: Returns number of sexes in the dataset
show: Prints out the PhenTestResult object
getCountMatrices: Returns count matrices if they are present (for methods "FE" and "RR"), NULL otherwise.

Natalja Kurbatova

    # Mixed Model framework
    file <- system.file("extdata", "test1.csv", package="PhenStat")
    test <- PhenStat:::PhenList(dataset=read.csv(file,na.strings = '-'),
            testGenotype="Sparc/Sparc")
    result <- PhenStat:::testDataset(test,
            equation="withoutWeight",
            depVariable="Bone.Area")
    PhenStat:::getVariable(result)
    PhenStat:::method(result)
    # Batch effect is significant
    PhenStat:::analysisResults(result)$model.effect.batch    
    # Variance homogeneosity 
    PhenStat:::analysisResults(result)$model.effect.variance
    # Weight effect is significant
    PhenStat:::analysisResults(result)$model.effect.weight     
    # Sex effect is significant 
    PhenStat:::analysisResults(result)$model.effect.interaction    
    # Sex effect p-value - the result of the test 
    PhenStat:::analysisResults(result)$model.output.interaction
    # Genotype effect p-value    
    PhenStat:::analysisResults(result)$model.output.genotype.nulltest.pVal
    # Final model formula with genotype
    PhenStat:::analysisResults(result)$model.formula.genotype 
    # Final model formula without genotype    
    PhenStat:::analysisResults(result)$model.formula.null 
    # Final model fitting output
    # result$model.output 
    # Final model fitting summary
    # summary(result$model.output)$tTable 
    
    # Fisher Exact Test framework
    file <- system.file("extdata", "test_categorical.csv", package="PhenStat")
    test <- PhenStat:::PhenList(dataset=read.csv(file,na.strings = '-'),
            testGenotype="Aff3/Aff3")
    result <- PhenStat:::testDataset(test,
            depVariable="Thoracic.Processes",
            method="FE") 
    PhenStat:::getVariable(result)
    PhenStat:::method(result)       
    for (i in seq_along(analysisResults(result))) {
        val <- PhenStat:::analysisResults(result)[[i]]
        val
    }