tests/testthat/test_TrainFastImputation.R

In FastImputation: Learn from Training Data then Quickly Fill in Missing Data

context("TrainFastImputation")

good_df <- data.frame(X1=letters, 
                      X2=1:26, 
                      X3=rnorm(26), 
                      X4=rexp(26), 
                      X5=runif(26), 
                      X6=sample(c(TRUE, FALSE), 26, replace=TRUE),
                      X7=3 - rexp(26))
good_constraints_upper <- list(list(7, list(upper=3)))
good_constraints_lower <- list(list(4, list(lower=0)))
good_constraints_both <- list(list(5, list(lower=0, upper=1)))
good_ignore_1 <- 1:2
good_ignore_2 <- c("X2", "X1")
good_categorical_1 <- 6
good_categorical_2 <- "X6"

warning_df <- data.frame(X3=rnorm(5), 
                         X6=sample(c(TRUE, FALSE), 5, replace=TRUE))
bad_array <- matrix(1:9, nrow=3)
bad_constraints_upper <- list(list(4, list(upper=0)))
bad_constraints_lower <- list(list(7, list(lower=4)))
bad_constraints_order <- list(list(5, list(lower=1, upper=0)))
bad_ignore_1 <- 10
bad_ignore_2 <- "X10"
bad_categorical_1 <- 10
bad_categorical_2 <- "X10"

test_that("TrainFastImputation catches bad input", {
  expect_error(TrainFastImputation(x=bad_array), 
               "'x' must be a dataframe.")
  expect_error(TrainFastImputation(x=good_df, constraints=bad_constraints_upper), 
               "Column 4 does not respect the upper bound specified.")
  expect_error(TrainFastImputation(x=good_df, constraints=bad_constraints_lower), 
               "Column 7 does not respect the lower bound specified.")
  expect_error(TrainFastImputation(x=good_df, constraints=bad_constraints_order), 
               "Column 5 does not respect the lower bound specified.")
  expect_error(TrainFastImputation(x=warning_df, categorical=good_categorical_2), 
               "Too few continuous variables.")
})

test_that("TrainFastImputation returns correct types", {
  res <- TrainFastImputation(x=good_df, 
                             constraints=good_constraints_both,
                             idvars=good_ignore_2)
  expect_true( is.list(res) )
  expect_true( is.character(res$FI_var_names) )
  expect_true( is.numeric(res$FI_means) )
  expect_true( is.vector(res$FI_means) )
  expect_true( det(res$FI_covariance) > 0 )
  expect_true( is.list(res$FI_constraints) )
  expect_true( identical(res$FI_cols_bound_to_intervals, 3L) )
  expect_true( is.numeric(res$FI_cols_categorical) )
})

test_that("TrainFastImputation returns correct values", {
  res_1 <- TrainFastImputation(x=good_df, 
                               constraints=good_constraints_upper, 
                               idvars=good_ignore_1, 
                               categorical=good_categorical_1)
  res_2 <- TrainFastImputation(x=good_df, 
                               constraints=good_constraints_lower, 
                               idvars=good_ignore_2, 
                               categorical=good_categorical_2)
  expect_equal( res_1$FI_var_names, paste("X", 1:7, sep="") )
  expect_equal( res_1$FI_constraints[[1]], list() )
  expect_equal( res_1$FI_constraints[[5]], list(upper=3) )
  expect_equal( res_1$FI_cols_to_ignore, c(1,2) )
  expect_equal( res_2$FI_cols_to_ignore, c(1,2) )
  expect_equal( res_1$FI_cols_categorical, 4 )
  expect_equal( res_2$FI_cols_categorical, 4 )
})