tests/testthat/test_base_wlda.r

context("wlda")

test_that("wlda: misspecified arguments", {
	data(iris)
	# wrong variable names
	expect_error(wlda(formula = Species ~ V1, data = iris))
	# wrong class
	expect_error(wlda(formula = iris, data = iris))
	# target variable also in x
	#expect_error(wlda(grouping = iris$Species, x = iris))          ## funktioniert, sollte aber nicht
	# missing x
	expect_error(wlda(grouping = iris$Species))
	## wrong method argument
	# missing quotes
	expect_error(wlda(Species ~ ., data = iris, method = ML))
	# method as vector
	expect_error(wlda(Species ~ ., data = iris, method = c("ML","unbiased")))
})


test_that("wlda throws a warning if grouping variable is numeric", {
	data(iris)
	expect_warning(wlda(formula = Petal.Width ~ ., data = iris))
	expect_warning(wlda(grouping = iris[,1], x = iris[,-1]))
	expect_warning(wlda(grouping = iris$Petal.Width, x = iris[,-5]))
})


test_that("wlda works if only one predictor variable is given", {
	data(iris)
	fit <- wlda(Species ~ Petal.Width, data = iris, subset = 6:60)
	expect_equal(ncol(fit$means), 1)	
	expect_equal(dim(fit$cov), rep(1, 2))	
})


test_that("wlda: training data from only one class", {
	data(iris)
	expect_that(wlda(Species ~ Petal.Width, data = iris, subset = 1:50), throws_error("training data from only one group given"))
	expect_that(wlda(Species ~ Petal.Width, data = iris, subset = 1), throws_error("training data from only one group given"))
})


test_that("wlda: weighting works correctly", {
	data(iris)
	## check if weighted solution with all weights = 1 equals unweighted solution
	fit1 <- wlda(Species ~ ., data = iris)
	fit2 <- wlda(Species ~ ., data = iris, weights = rep(1,150))
	expect_equal(fit1[-9],fit2[-9])
	## returned weights	
	a <- rep(1,150)
	names(a) <- 1:150
	expect_equal(fit1$weights, a)
	expect_equal(fit2$weights, a)
	## weights and subsetting
	# formula, data
	fit <- wlda(Species ~ ., data = iris, subset = 11:60)
	a <- rep(1,50)
	names(a) <- 11:60
	expect_equal(fit$weights, a)
	# formula, data, weights
	fit <- wlda(Species ~ ., data = iris, weights = rep(1:3, 50), subset = 11:60)
	b <- rep(1:3,50)[11:60]
	names(b) <- 11:60
	expect_equal(fit$weights, b)
	# x, grouping
	fit <- wlda(x = iris[,-5], grouping = iris$Species, subset = 11:60)
	expect_equal(fit$weights, a)	
	# x, grouping, weights
	fit <- wlda(x = iris[,-5], grouping = iris$Species, weights = rep(1:3, 50), subset = 11:60)
	expect_equal(fit$weights, b)
	## wrong specification of weights argument
	# weights in a matrix
	weight <- matrix(seq(1:150),nrow=50)
	expect_error(wlda(Species ~ ., data = iris, weights = weight))
	# weights < 0
	expect_error(wlda(Species ~ ., data = iris, weights = rep(-5, 150)))
	# weights true/false
	expect_error(wlda(Species ~ ., data = iris, weights = TRUE))
})


test_that("wlda: subsetting works", {
	data(iris)
	# formula, data
	fit1 <- wlda(Species ~ ., data = iris, subset = 1:80)
	fit2 <- wlda(Species ~ ., data = iris[1:80,])
	expect_equal(fit1[-9],fit2[-9])
	# formula, data, weights
	fit1 <- wlda(Species ~ ., data = iris, weights = rep(1:3, each = 50), subset = 1:80)
	fit2 <- wlda(Species ~ ., data = iris[1:80,], weights = rep(1:3, each = 50)[1:80])
	expect_equal(fit1[-9],fit2[-9])
	expect_equal(length(fit1$weights), 80)
	# x, grouping
	fit1 <- wlda(grouping = iris$Species, x = iris[,-5], subset = 1:80)
	fit2 <- wlda(grouping = iris$Species[1:80], x = iris[1:80,-5])
	expect_equal(fit1[c(1:8)],fit2[c(1:8)])
	# x, grouping, weights
	fit1 <- wlda(grouping = iris$Species, x = iris[,-5], weights = rep(1:3, each = 50), subset = 1:80)
	fit2 <- wlda(grouping = iris$Species[1:80], x = iris[1:80,-5], weights = rep(1:3, each = 50)[1:80])
	expect_equal(fit1[c(1:8)],fit2[c(1:8)])
	expect_equal(length(fit1$weights), 80)
	# wrong specification of subset argument
	expect_error(wlda(Species ~ ., data = iris, subset = iris[1:10,]))
	expect_error(wlda(Species ~ ., data = iris, subset = FALSE))
	expect_error(wlda(Species ~ ., data = iris, subset = 0))
	expect_error(wlda(Species ~ ., data = iris, subset = -10:50))
})


test_that("wlda: NA handling works correctly", {
	### NA in x
	data(iris)
	irisna <- iris
	irisna[1:10, c(1,3)] <- NA
	## formula, data
	# na.fail
	expect_error(wlda(Species ~ ., data = irisna, subset = 6:60, na.action = na.fail))
	# check if na.omit works correctly
	fit1 <- wlda(Species ~ ., data = irisna, subset = 6:60, na.action = na.omit)
	fit2 <- wlda(Species ~ ., data = irisna, subset = 11:60)
	expect_equal(fit1[-c(9, 12)], fit2[-9])
	expect_equal(length(fit1$weights), 50)
	## formula, data, weights
	# na.fail
	expect_error(wlda(Species ~ ., data = irisna, subset = 6:60, weights = rep(1:3, 50), na.action = na.fail))
	# check if na.omit works correctly
	fit1 <- wlda(Species ~ ., data = irisna, subset = 6:60, weights = rep(1:3, 50), na.action = na.omit)
	fit2 <- wlda(Species ~ ., data = irisna, subset = 11:60, weights = rep(1:3, 50))
	expect_equal(fit1[-c(9, 12)], fit2[-9])
	expect_equal(length(fit1$weights), 50)
	## x, grouping
	# na.fail
	expect_error(wlda(grouping = irisna$Species, x = irisna[,-5], subset = 6:60, na.action = na.fail))
	# check if na.omit works correctly
	fit1 <- wlda(grouping = irisna$Species, x = irisna[,-5], subset = 6:60, na.action = na.omit)
	fit2 <- wlda(grouping = irisna$Species, x = irisna[,-5], subset = 11:60)
	expect_equal(fit1[1:8], fit2[1:8])
	expect_equal(length(fit1$weights), 50)
	## x, grouping, weights
	# na.fail
	expect_error(wlda(grouping = irisna$Species, x = irisna[,-5], subset = 6:60, weights = rep(1:3, 50), na.action = na.fail))
	# check if na.omit works correctly
	fit1 <- wlda(grouping = irisna$Species, x = irisna[,-5], subset = 6:60, weights = rep(1:3, 50), na.action = na.omit)
	fit2 <- wlda(grouping = irisna$Species, x = irisna[,-5], subset = 11:60, weights = rep(1:3, 50))
	expect_equal(fit1[1:8], fit2[1:8])
	expect_equal(length(fit1$weights), 50)
	
	### NA in grouping
	irisna <- iris
	irisna$Species[1:10] <- NA
	## formula, data
	# na.fail
	expect_error(wlda(Species ~ ., data = irisna, subset = 6:60, na.action = na.fail))
	# check if na.omit works correctly
	fit1 <- wlda(Species ~ ., data = irisna, subset = 6:60, na.action = na.omit)
	fit2 <- wlda(Species ~ ., data = irisna, subset = 11:60)
	expect_equal(fit1[-c(9, 12)], fit2[-9])
	expect_equal(length(fit1$weights), 50)
	## formula, data, weights
	# na.fail
	expect_error(wlda(Species ~ ., data = irisna, subset = 6:60, weights = rep(1:3, 50), na.action = na.fail))
	# check if na.omit works correctly
	fit1 <- wlda(Species ~ ., data = irisna, subset = 6:60, weights = rep(1:3, 50), na.action = na.omit)
	fit2 <- wlda(Species ~ ., data = irisna, subset = 11:60, weights = rep(1:3, 50))
	expect_equal(fit1[-c(9, 12)], fit2[-9])
	expect_equal(length(fit1$weights), 50)
	## x, grouping
	# na.fail
	expect_error(wlda(grouping = irisna$Species, x = irisna[,-5], subset = 6:60, na.action = na.fail))
	# check if na.omit works correctly
	fit1 <- wlda(grouping = irisna$Species, x = irisna[,-5], subset = 6:60, na.action = na.omit)
	fit2 <- wlda(grouping = irisna$Species, x = irisna[,-5], subset = 11:60)
	expect_equal(fit1[1:8], fit2[1:8])
	expect_equal(length(fit1$weights), 50)
	## x, grouping, weights
	# na.fail
	expect_error(wlda(grouping = irisna$Species, x = irisna[,-5], subset = 6:60, weights = rep(1:3, 50), na.action = na.fail))
	# check if na.omit works correctly
	fit1 <- wlda(grouping = irisna$Species, x = irisna[,-5], subset = 6:60, weights = rep(1:3, 50), na.action = na.omit)
	fit2 <- wlda(grouping = irisna$Species, x = irisna[,-5], subset = 11:60, weights = rep(1:3, 50))
	expect_equal(fit1[1:8], fit2[1:8])
	expect_equal(length(fit1$weights), 50)

	### NA in weights
	weights <- rep(1:3,50)
	weights[1:10] <- NA
	## formula, data, weights
	# na.fail
	expect_error(wlda(Species ~ ., data = iris, subset = 6:60, weights = weights, na.action = na.fail))
	# check if na.omit works correctly
	fit1 <- wlda(Species ~ ., data = iris, subset = 6:60, weights = weights, na.action = na.omit)
	fit2 <- wlda(Species ~ ., data = iris, subset = 11:60, weights = weights)
	expect_equal(fit1[-c(9, 12)], fit2[-9])
	expect_equal(length(fit1$weights), 50)
	## x, grouping, weights
	# na.fail
	expect_error(wlda(grouping = iris$Species, x = iris[,-5], subset = 6:60, weights = weights, na.action = na.fail))
	# check if na.omit works correctly
	fit1 <- wlda(grouping = iris$Species, x = iris[,-5], subset = 6:60, weights = weights, na.action = na.omit)
	fit2 <- wlda(grouping = iris$Species, x = iris[,-5], subset = 11:60, weights = weights)
	expect_equal(fit1[1:8], fit2[1:8])
	expect_equal(length(fit1$weights), 50)

	### NA in subset
	subset <- 6:60
	subset[1:5] <- NA
	## formula, data
	# na.fail
	expect_error(wlda(Species ~ ., data = iris, subset = subset, na.action = na.fail))
	# check if na.omit works correctly
	fit1 <- wlda(Species ~ ., data = iris, subset = subset, na.action = na.omit)
	fit2 <- wlda(Species ~ ., data = iris, subset = 11:60)
	expect_equal(fit1[-c(9, 12)], fit2[-9])
	expect_equal(length(fit1$weights), 50)
	## formula, data, weights
	# na.fail
	expect_error(wlda(Species ~ ., data = iris, subset = subset, weights = rep(1:3, 50), na.action = na.fail))
	# check if na.omit works correctly
	fit1 <- wlda(Species ~ ., data = iris, subset = subset, weights = rep(1:3, 50), na.action = na.omit)
	fit2 <- wlda(Species ~ ., data = iris, subset = 11:60, weights = rep(1:3, 50))
	expect_equal(fit1[-c(9, 12)], fit2[-9])
	expect_equal(length(fit1$weights), 50)
	## x, grouping
	# na.fail
	expect_error(wlda(grouping = iris$Species, x = iris[,-5], subset = subset, na.action = na.fail))
	# check if na.omit works correctly
	fit1 <- wlda(grouping = iris$Species, x = iris[,-5], subset = subset, na.action = na.omit)
	fit2 <- wlda(grouping = iris$Species, x = iris[,-5], subset = 11:60)
	expect_equal(fit1[1:8], fit2[1:8])
	expect_equal(length(fit1$weights), 50)
	## x, grouping, weights
	# na.fail
	expect_error(wlda(grouping = iris$Species, x = iris[,-5], subset = subset, weights = rep(1:3, 50), na.action = na.fail))
	# check if na.omit works correctly
	fit1 <- wlda(grouping = iris$Species, x = iris[,-5], subset = subset, weights = rep(1:3, 50), na.action = na.omit)
	fit2 <- wlda(grouping = iris$Species, x = iris[,-5], subset = 11:60, weights = rep(1:3, 50))
	expect_equal(fit1[1:8], fit2[1:8])
	expect_equal(length(fit1$weights), 50)
})


#=================================================================================================================
context("predict.wlda")

test_that("predict.wlda works correctly with formula and data.frame interface and with missing newdata", {
	data(iris)
	ran <- sample(1:150,100)
	## formula, data
	fit <- wlda(formula = Species ~ ., data = iris, subset = ran)  
  	pred <- predict(fit)
  	expect_equal(rownames(pred$posterior), rownames(iris)[ran])  	
	## formula, data, newdata
	fit <- wlda(formula = Species ~ ., data = iris, subset = ran)  
  	predict(fit, newdata = iris[-ran,])
	## grouping, x
	fit <- wlda(x = iris[,-5], grouping = iris$Species, subset = ran)  
  	pred <- predict(fit)
  	expect_equal(rownames(pred$posterior), rownames(iris)[ran])  	
	## grouping, x, newdata
	fit <- wlda(x = iris[,-5], grouping = iris$Species, subset = ran)  
  	predict(fit, newdata = iris[-ran,-5])
})


test_that("predict.wlda: retrieving training data works", {
	data(iris)
	## no subset
	# formula, data
	fit <- wlda(formula = Species ~ ., data = iris)
  	pred1 <- predict(fit)
  	pred2 <- predict(fit, newdata = iris)
  	expect_equal(pred1, pred2)
	# y, x
	fit <- wlda(x = iris[,-5], grouping = iris$Species)  
  	pred1 <- predict(fit)
  	pred2 <- predict(fit, newdata = iris[,-5])
  	expect_equal(pred1, pred2)
	## subset
	ran <- sample(1:150,100)
	# formula, data
	fit <- wlda(formula = Species ~ ., data = iris, subset = ran)
  	pred1 <- predict(fit)
  	pred2 <- predict(fit, newdata = iris[ran,])
  	expect_equal(pred1, pred2)
	# y, x
	fit <- wlda(x = iris[,-5], grouping = iris$Species, subset = ran)  
  	pred1 <- predict(fit)
  	pred2 <- predict(fit, newdata = iris[ran,-5])
  	expect_equal(pred1, pred2)
})


test_that("predict.wlda works with missing classes in the training data", {
	data(iris)
	ran <- sample(1:150,100)
	expect_warning(fit <- wlda(Species ~ ., data = iris, subset = 1:100))
	pred <- predict(fit, newdata = iris[-ran,])
	expect_equal(nlevels(pred$class), 3)
	expect_equal(ncol(pred$posterior), 2)
	# a <- rep(0,50)
	# names(a) <- rownames(pred$posterior)
	# expect_equal(pred$posterior[,3], a)
})


test_that("predict.wlda works with one single predictor variable", {
	data(iris)
	ran <- sample(1:150,100)
	fit <- wlda(Species ~ Petal.Width, data = iris, subset = ran)
	expect_equal(ncol(fit$means), 1)
	expect_equal(dim(fit$cov), rep(1, 2))
	predict(fit, newdata = iris[-ran,])
})


test_that("predict.wlda works with one single test observation", {
	data(iris)
	ran <- sample(1:150,100)
	fit <- wlda(Species ~ Petal.Width, data = iris, subset = ran)
  	pred <- predict(fit, newdata = iris[5,])
	expect_equal(length(pred$class), 1)
	expect_equal(dim(pred$posterior), c(1, 3))
	a <- factor("setosa", levels = c("setosa", "versicolor", "virginica"))
	names(a) = "5"
	expect_equal(pred$class, a)
	pred <- predict(fit, newdata = iris[58,])
	expect_equal(length(pred$class), 1)
	expect_equal(dim(pred$posterior), c(1, 3))
	a <- factor("versicolor", levels = c("setosa", "versicolor", "virginica"))
	names(a) = "58"
	expect_equal(pred$class, a)
})	


test_that("predict.wlda works with one single predictor variable and one single test observation", {
	data(iris)
	ran <- sample(1:150,100)
	fit <- wlda(Species ~ Petal.Width, data = iris, subset = ran)
	expect_equal(ncol(fit$means), 1)
	expect_equal(dim(fit$cov), rep(1, 2))
	pred <- predict(fit, newdata = iris[5,])
	expect_equal(length(pred$class), 1)
	expect_equal(dim(pred$posterior), c(1, 3))
})

   
test_that("predict.wlda: NA handling in newdata works", {
	data(iris)
	ran <- sample(1:150,100)
	irisna <- iris
	irisna[1:17,c(1,3)] <- NA
	fit <- wlda(Species ~ ., data = iris, subset = ran)
	expect_warning(pred <- predict(fit, newdata = irisna))
	expect_equal(all(is.na(pred$class[1:17])), TRUE)
	expect_equal(all(is.na(pred$posterior[1:17,])), TRUE)	
})


test_that("predict.wlda: misspecified arguments", {
	data(iris)
	ran <- sample(1:150,100)
	fit <- wlda(Species ~ Petal.Width, data = iris, subset = ran)
    # errors in newdata
    expect_error(predict(fit, newdata = TRUE))
    expect_error(predict(fit, newdata = -50:50))
    # errors in prior
    expect_error(predict(fit, prior = rep(2,length(levels(iris$Species))), newdata = iris[-ran,]))
    expect_error(predict(fit, prior = TRUE, newdata = iris[-ran,]))
    expect_error(predict(fit, prior = 0.6, newdata = iris[-ran,]))
})
schiffner/locClass documentation built on May 29, 2019, 3:39 p.m.