tests/testthat/testRai.R

In korydjohnson/rai: Revisiting-Alpha-Investing for Polynomial Regression

context("rai Wrapper")
library(rai)

# set up ------------------------------------------------------------------
data("CO2")

# tests -------------------------------------------------------------------
test_that("no errors", {
  # happens on step 12 of the following data example (colinearity)
  data("mtcars")
  theResponse = mtcars$mpg
  theData = mtcars[ ,-1]
  rai_out = rai(theData, theResponse)
  rai_out = rai(theData, theResponse, alg="raiPlus")
})

test_that("prepareData for numeric matrix", {
  theData = matrix(1:10, nrow=2)
  theData_mm = prepareData(theData)
  attributes(theData_mm)$dimnames = NULL
  expect_identical(theData, theData_mm)
})

test_that("prepareData for data.frame", {
  theData = head(CO2)
  theData_mm = prepareData(theData)
  theData_mm2 = model.matrix(~.-1, data=theData)
  redundantCols = apply(theData_mm2, 2, function(col) length(unique(col)))
  theData_mm2 = theData_mm2[, redundantCols>1]
  expect_identical(theData_mm, theData_mm2)
})

test_that("modMissing works", {
  theData = CO2
  theData[1,1] = NA
  theData[1,5] = NA
  theData_mm = prepareData(theData)
  expect_equal(ncol(theData_mm), 18)
  expect_equal(theData_mm[1,"uptake"], mean(theData[-1,"uptake"]))
  expect_equal(unname(theData_mm[,"PlantNA"]), c(1, rep(0, 83)))
})

test_that("categorical data converted; pass & epoch", {
  theResponse = CO2$uptake
  theData = CO2[ ,-5]
  rai_out = rai(theData, theResponse)
  rai_out = rai(theData, theResponse, alg="raiPlus", verbose=T)
})

test_that("1 column output for binary categories, n col for n > 2 categories", {
  theData = data.frame("binary1" = as.character(c(1,2,1)),
                       "binary2" = as.character(c(2,1,1)),
                       "multi1" = as.character(1:3),
                       "multi2" = as.character(4:6))
  prepData = prepareData(theData)
  expect_identical(length(unique(colnames(prepData))), ncol(prepData))
  expect_identical(ncol(prepData), 8L)
  prepData = prepareData(theData[,c(1,2)])
  expect_identical(ncol(prepData), 2L)
  prepData = prepareData(theData[,c(3,4)])
  expect_identical(ncol(prepData), 6L)
})

test_that("baseModel works", {
  data("mtcars")
  theResponse = mtcars$mpg
  theData = mtcars[ ,-1]
  colnames(theData)
  baseModel = list("cyl", "hp")
  baseIndexes = match(baseModel, colnames(theData))
  rai_out = rai(theData, theResponse, baseModel=baseModel)
  expect_equal(unlist(rai_out$features[1:length(baseModel)]),
               baseIndexes)
  baseModel = list(c("cyl","hp"))
  rai_out = rai(theData, theResponse, baseModel=baseModel)
  expect_true(grepl("cyl\\*hp", rai_out$formula))
  rai_out = rai(theData, theResponse, baseModel=baseModel)
})