barPlotModelComparison-ClinicalModel-PCOSP_or_RLS_or_RGA-method: Make a Bar Plot Comparison Model Performance Between a...

In bhklab/PDATK: Pancreatic Ductal Adenocarcinoma Tool-Kit

Description

Make a Bar Plot Comparison Model Performance Between a ClinicalModel and a PCOSP, RLSModel or RGAModel object.

Usage

## S4 method for signature 'ClinicalModel,PCOSP_or_RLS_or_RGA'
barPlotModelComparison(model1, model2, stat, ...)

Arguments

model1	A ClinicalModel object.
model2	A PCOSP or RLSModel or RGAModel object.
stat	A character vector specifying which statistic to compare the models using. Options are 'AUC', 'log_D_index' or 'concordance_index'.
...	Not used.

Value

A ggplot2 object showing a barplot coloured by the model and comparing the stat between all cohorts that both models were validated with.

Examples

data(sampleValPCOSPmodel)
data(sampleCohortList)
data(sampleICGCmicro)

# Set parallelization settings
BiocParallel::register(BiocParallel::SerialParam())

# Setup the models
clinicalModel <- ClinicalModel(sampleICGCmicro,
  formula='prognosis ~ sex + age + T + N + M + grade',
  randomSeed=1987)

# Train the models
trainedClinicalModel <- trainModel(clinicalModel)

# Make predctions
clinicalPredCohortList <- predictClasses(sampleCohortList[c('PCSI', 'TCGA')],
  model=trainedClinicalModel)

# Validate the models
validatedClinicalModel <- validateModel(trainedClinicalModel,
  valData=clinicalPredCohortList)

# Plot the comparison
modelCompBarPlot <- barPlotModelComparison(validatedClinicalModel,
 sampleValPCOSPmodel, stat='AUC')

bhklab/PDATK documentation built on Dec. 27, 2021, 7:46 a.m.