tests/testthat/test_mob_lda.R

In schiffner/locClass: Collection of Local Classification Methods

## mob
# 1. fit global model -> errors in global fit result in an exception, mob tree does not exist
# main recursion
# 2. mob_fit: -> reweight -> reweight can return an object of class try-error, in this case the node is empty and the respective branch of the tree stops to grow
# 3. mob_fit_setupnode: 
#		try(mob_fit_fluctests): estfun (estfun cannot return an object of class try-error) -> errors in fluctuation tests result in a test statistic with value NA, the respective node is not split and the corresponding branch stops to grow
#		mob_fit_splitnode: calls for every candidate split point: mob_fit_getobjfun: reweight, objfun (deviance) -> 
#							reweight can return an object of class try-error, for this reason objfun(try-error) should return an infinite value
#							if an error occurs in objfun itself it should return an infinite value
# 4. mob_fit_childweights
# 5. mob_fit ...


context("mobLdaModel")

test_that("mobLdaModel: misspecified arguments", {
	# wrong variable names
	expect_that(mob(Species ~ V1 | Sepal.Length, data = iris, model = ldaModel,
		control = mob_control(objfun = deviance, minsplit = 20)), throws_error("object 'V1' not found"))
	expect_that(mob(Species ~ Sepal.Length | V1, data = iris, model = ldaModel,
		control = mob_control(objfun = deviance, minsplit = 20)), throws_error("object 'V1' not found"))
	expect_that(mob(y ~ Sepal.Length | Sepal.Width, data = iris, model = ldaModel,
		control = mob_control(objfun = deviance, minsplit = 20)), throws_error("object 'y' not found"))
	# wrong class
	expect_error(mob(iris, data = iris, model = ldaModel,
		control = mob_control(objfun = deviance, minsplit = 20)))
	# target variable also in x
	# expect_error(mob(Species ~ Species + Sepal.Length | Sepal.Width, data = iris, model = ldaModel,
		# control = mob_control(objfun = deviance, minsplit = 20)))	## funktioniert, sollte aber nicht
})


test_that("mobLdaModel: binary problem", {
	data <- benchData::vData(500)
	fit <- mob(y ~ x.1 + x.2 | x.1 + x.2, data = data, model = ldaModel,
		control = mob_control(objfun = deviance, minsplit = 200))
	tr <- mean(predict(fit) != data$y)
	ba <- mean(benchData::vBayesClass(data$x) != data$y)
	expect_true(tr < ba + 0.05)	
})


test_that("mobLdaModel: multi-class problem", {
	data <- benchData::xor3Data(1000)
	fit <- mob(y ~ x.1 + x.2 | x.1 + x.2, data = data, model = ldaModel,
		control = mob_control(objfun = deviance, minsplit = 100))
	tr <- mean(predict(fit) != data$y)
	ba <- mean(benchData::xor3BayesClass(data$x) != data$y)
	expect_true(tr < ba + 0.05)	
})


test_that("mobLdaModel throws a warning if grouping variable is numeric", {
	expect_that(fit <- mob(Petal.Width ~ . | Sepal.Length, data = iris, model = ldaModel,
		control = mob_control(objfun = deviance, minsplit = 20)), gives_warning("'grouping' was coerced to a factor"))
	### try-error in estfun.wlda: covariance is singular
})


test_that("mobLdaModel works if only one predictor variable is given", {
	fit <- mob(Species ~ Sepal.Width | Sepal.Length, data = iris, model = ldaModel,
		control = mob_control(objfun = deviance, minsplit = 20))
	## warnings about empty classes in some nodes
	terminal <- nodes(fit, max(where(fit)))
	expect_equal(ncol(terminal[[1]]$model$means), 1)
	expect_equal(dim(terminal[[1]]$model$cov), c(1,1))
})


test_that("mobLdaModel: Local and global solution coincide if minsplit is large", {
	data <- benchData::vData(500)
	fit <- mob(y ~ x.1 + x.2 | x.1 + x.2, data = data, model = ldaModel,
		control = mob_control(objfun = deviance, minsplit = 500))
	w <- wlda(y ~ ., data = as.data.frame(data), method = "ML")
	expect_equal(fit@tree$model$prior, w$prior)
	expect_equal(fit@tree$model$means, w$means)
	expect_equal(fit@tree$model$cov, w$cov)
	pred <- predict(fit)
	p <- predict(w)
	expect_equal(pred, as.numeric(p$class))
})


test_that("mobLdaModel: training data from only one class", {
	expect_that(fit <- mob(Species ~ Sepal.Width | Sepal.Length, data = iris[1:50,], model = ldaModel,
		control = mob_control(objfun = deviance, minsplit = 20)), throws_error("training data from only one group given"))
	## error in global fit
})


test_that("mobLdaModel: arguments are passed to wlda",{
	fit <- mob(Species ~ Sepal.Width | Sepal.Length, data = iris, model = ldaModel,
		control = mob_control(objfun = deviance, minsplit = 20))
	# default: "ML" hard-coded
	expect_equal(fit@tree$right$model$method, "ML")	
	# unbiased
	expect_error(fit <- mob(Species ~ Sepal.Width | Sepal.Length, data = iris, model = ldaModel, method = "unbiased",
		control = mob_control(objfun = deviance, minsplit = 20)))
})


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

test_that("predict.ldaModel works correctly with formula interface and with missing newdata", {
	ran <- sample(1:150,100)
	## formula, data
	fit <- mob(Species ~ . | Sepal.Length, data = iris[ran,], model = ldaModel,
		control = mob_control(objfun = deviance, minsplit = 20))	
  	pred <- predict(fit)
	mean(pred != as.numeric(iris$Species[ran]))
  	pred <- predict(fit, out = "posterior")
  	p <- matrix(0, length(pred), 3)
	colnames(p) = levels(iris$Species)
	rownames(p) = sapply(pred, rownames)
	for (i in seq_along(pred)) {
		p[i, colnames(pred[[i]])] = pred[[i]]
	}
  	expect_equal(rownames(p), rownames(iris)[ran])  	
	## formula, data, newdata
	p <- predict(fit, newdata = iris[-ran,])
	mean(p != as.numeric(iris$Species[-ran]))
})


test_that("predict.ldaModel: retrieving training data works", {
	## no subset
	# formula, data
	fit <- mob(Species ~ . | Sepal.Length, data = iris, model = ldaModel,
		control = mob_control(objfun = deviance, minsplit = 20))	
  	pred1 <- predict(fit)
  	pred2 <- predict(fit, newdata = iris)
  	expect_equal(pred1, pred2)
	## subset
	ran <- sample(1:150,100)
	# formula, data
	fit <- mob(Species ~ . | Sepal.Length, data = iris[ran,], model = ldaModel,
		control = mob_control(objfun = deviance, minsplit = 20))	
  	pred1 <- predict(fit)
  	pred2 <- predict(fit, newdata = iris[ran,])
  	expect_equal(pred1, pred2)
})


test_that("predict.ldaModel works with missing classes in the training data", {
	ran <- sample(1:150,100)
	fit <- mob(Species ~ . | Sepal.Length, data = iris[1:100,], model = ldaModel,
		control = mob_control(objfun = deviance, minsplit = 20))
	## sporadic try-errors in reweight: training data from only one group 
	pred <- predict(fit, newdata = iris[-ran,])
	pred <- predict(fit, newdata = iris[-ran,], out = "posterior")
	# expect_equal(nlevels(pred$class), 3)
	# expect_equal(ncol(pred$posterior), 2)
})


test_that("predict.ldaModel works with one single predictor variable", {
	ran <- sample(1:150,100)
	fit <- mob(Species ~ Sepal.Width | Sepal.Width, data = iris[ran,], model = ldaModel,
		control = mob_control(objfun = deviance, minsplit = 20))
	terminal <- nodes(fit, max(where(fit)))
	expect_equal(ncol(terminal[[1]]$model$means), 1)
	expect_equal(dim(terminal[[1]]$model$cov), c(1,1))
	predict(fit, newdata = iris[-ran,])
})


test_that("predict.ldaModel works with one single test observation", {
	ran <- sample(1:150,100)
	fit <- mob(Species ~ . | Sepal.Width, data = iris[ran,], model = ldaModel,
		control = mob_control(objfun = deviance, minsplit = 20))
  	pred <- predict(fit, newdata = iris[5,])
	expect_equal(length(pred), 1)
  	pred <- predict(fit, newdata = iris[5,], out = "posterior")
	expect_equal(dim(pred[[1]]), c(1, 3))
})	


test_that("predict.ldaModel: NA handling in newdata works", {
	## NAs in explanatory variables are ok
	ran <- sample(1:150,100)
	irisna <- iris
	irisna[1:17,c(1,3)] <- NA
	fit <- mob(Species ~ . | Sepal.Width, data = iris[ran,], model = ldaModel,
		control = mob_control(objfun = deviance, minsplit = 30))
	expect_warning(pred <- predict(fit, newdata = irisna))
	## warnings: kein nicht-fehlendes Argument für max; gebe -Inf zurück
	expect_true(all(is.na(pred[1:17])))

	## NAs in splitting variable are not ok
	irisna[1:17,1:3] <- NA
	fit <- mob(Species ~ . | Sepal.Width, data = iris[ran,], model = ldaModel,
		control = mob_control(objfun = deviance, minsplit = 30))
	expect_error(pred <- predict(fit, newdata = irisna)) 
	## error: VECTOR_ELT() can only be applied to a 'list', not a 'NULL'
})


test_that("predict.ldaModel: misspecified arguments", {
	ran <- sample(1:150,100)
	fit <- mob(Species ~ . | Sepal.Width, data = iris[ran,], model = ldaModel,
		control = mob_control(objfun = deviance, minsplit = 20))	
    # errors in newdata
    expect_error(predict(fit, newdata = TRUE))
    expect_error(predict(fit, newdata = -50:50))
})