tests/testthat/test-impute_median.R

In davidbiol/empire: Estimation of Missing Data Using Penalized Iterative Regressions

## data
x1 <- c(1.30, 1.40, NA, 1.54, 1.30, NA, 1.50, 1.60, 1.41, 1.34)
x2 <- c(51, 60, 69, 73, 56, 75, 80, NA, 58, 70)
x3 <- c(4.8, NA, 5.8, 6.5, 5.3, 7.0, 8.1, 7.8, 5.9, 6.1)
x4 <- c(115, 130, 138, 148, 122, 152, 160, 155, 135, 140)
df <- data.frame(x1,x2,x3,x4)

## Check it is accurately estimating

test_that("Values are accurately estimated", {
  object_median <- empire::impute_median(df)
  median_test <- vector()
  for (i in seq_len(ncol(df))) {
    var_sort <- sort(na.omit(df[,i]))
    if (!length(var_sort)%%2){ # It is even
      median_test[i] <- mean(c(var_sort[length(var_sort)/2], var_sort[(length(var_sort)/2) + 1]))
    }
    else { # It is odd
      median_test[i] <- var_sort[ceiling(length(var_sort)/2)]
    }
  }
  expect_equal(as.vector(sapply(df, stats::median, na.rm=TRUE)), median_test) # Median is right calculated with stats::median()
  expect_equal(object_median$imp_values, c(1.405, 1.405, 69.000, 6.100))
})
## Error testing
df_list <- list(x1 = x1, x2 = x2, x3 = x3, x4 = x4)

test_that("Error if data input is not a data frame or a matrix", {
  expect_error(empire::impute_median(df_list))
})

x5 <- c(117, NA, 118, 128, 145, 161, 170, 155, "hello", 140)
df_with_character <- data.frame(x1,x2,x3,x4,x5)
df_matrix_with_character <- matrix(c(x1,x2,x3,x4,x5), ncol = 5)

test_that("Error if data input is a vector of characters", {
  expect_error(empire::impute_median(df_with_character))
  expect_error(empire::impute_median(df_matrix_with_character))
})