inst/Runit/runit.performAnalysis.R

In MSToolkit: The MSToolkit library for clinical trial design

test.convertSASIncludeRows <- function() {
	
	checkTrue(convertSASIncludeRows() == "**** No doses to drop ****;")
	checkException(convertSASIncludeRows(1:5))
	checkException(convertSASIncludeRows(rbind(1:5)))

	inMat <- cbind(1:3, c(0, 15, 30)) 
	str1 <- convertSASIncludeRows( inMat, doseCol = "DOSECOL", interimCol = "INTERIMCOL")
	target1 <- "IF ( INTERIMCOL = 1 and DOSECOL = 0 ) or ( INTERIMCOL = 2 and DOSECOL = 15 ) or ( INTERIMCOL = 3 and DOSECOL = 30 ) ;"
	checkTrue(str1 == target1)
	
}

test.performAnalysis <- function( ){
	
	# Example data
	myDf <- data.frame(SUBJ = 1:80, .dose = rep(c(0, 15, 30, 45), each = 20))
	myDf$.interim <- rep(1:2, 40)
	myDf$RESP <- log(with(myDf, abs(exp(sqrt(.dose) + 1 + rnorm(nrow(myDf))))))
	
	# Analysis code
	lmCode <- function(data) {
		myLm <- lm(RESP ~ .dose, data = data)
		newData <- data.frame(.dose = sort(unique(data$.dose)))
		doseMeans <- as.data.frame(predict(myLm, newData, se.fit = TRUE, interval = "confidence")$fit)
		names(doseMeans) <- c("Mean", "Lower", "Upper")
		data.frame(.dose = sort(unique(data$.dose)), doseMeans, N = as.vector(table(data$.dose)))
	}
	
	# Perform analysis
	out1 <- performAnalysis(lmCode, data = myDf, doseCol = ".dose", interimCol = ".interim")
	
	# Tests
	checkTrue(is.data.frame(out1) & all(names(out1) == c(".dose", "Mean", "Lower", "Upper", "N")))
	checkTrue(all(out1$N == 20))

	# Allow analysis of a single dataset given R Script reference
	# Example data
	myDf <- data.frame(SUBJ = 1:80, DOSE = rep(c(0, 15, 30, 45), each = 20))
	myDf$INTERIM <- rep(1:2, 40)
	myDf$RESP <- log(with(myDf, abs(exp(sqrt(DOSE) + 1 + rnorm(nrow(myDf))))))
	
	# Perform analysis
	whichPath <- system.file(package = "MSToolkit", "systemTest", "data", "Scripts")
	out1 <- performAnalysis("rAnalysisScript.R", data = myDf, workingPath = whichPath)
	
	# Tests
	checkTrue(is.data.frame(out1) & all(names(out1) == c("DOSE", "Mean", "Lower", "Upper", "N")))
	checkTrue(all(out1$N == 20))

	# Allow subsets and data steps to be made to the data before analysing
	# Example data
	myDf <- data.frame(SUBJ = 1:80, DOSE = rep(c(0, 15, 30, 45), each = 20))
	myDf$INTERIM <- rep(1:2, 40)
	myDf$RESP <- log(with(myDf, abs(exp(sqrt(DOSE) + 1 + rnorm(nrow(myDf))))))
	
	# Analysis code
	lmCode <- function(data) {
		myLm <- lm(RESP ~ DOSE, data = data)
		newData <- data.frame(DOSE = sort(unique(data$DOSE)))
		doseMeans <- as.data.frame(predict(myLm, newData, se.fit = TRUE, interval = "confidence")$fit)
		names(doseMeans) <- c("Mean", "Lower", "Upper")
		data.frame(DOSE = sort(unique(data$DOSE)), doseMeans, N = as.vector(table(data$DOSE)))
	}
	
	# Perform analysis
	out1 <- performAnalysis(lmCode, data = myDf,
			includeRows = cbind(c(1, 1, 1, 2, 2, 2), c(0, 15, 30, 0, 30, 45)))
	
	# Tests
	checkTrue(is.data.frame(out1) & all(names(out1) == c("DOSE", "Mean", "Lower", "Upper", "N")))
	checkTrue(all(out1$N == rep(c(20, 10), 2)))

	# Allow checking of return structure against
	# expected "Micro evaluation" return format
	myDf <- data.frame(SUBJ = 1:80, .dose = rep(c(0, 15, 30, 45), each = 20))
	myDf$.interim <- rep(1:2, 40)
	myDf$RESP <- log(with(myDf, abs(exp(sqrt(.dose) + 1 + rnorm(nrow(myDf))))))
	
	# Analysis code
	aCode1 <- function(data) 1:5
	aCode2 <- function(data) data.frame(X = 1:5)
	
	# Perform analysis
	out1 <- try(performAnalysis(aCode1, data = myDf), silent = TRUE)
	out2 <- try(performAnalysis(aCode2, data = myDf), silent = TRUE)
	checkTrue(class(out1) == "try-error")
	checkTrue(is.data.frame(out2))

}