Improvement of datacleaner package

NAMESPACE

@@ -1 +1,6 @@
-exportPattern("^[[:alpha:]]+")
+# Generated by roxygen2: do not edit by hand
+
+export(meanimpute)
+export(transform_log)
+export(windsorize)
+import(stats)

R/meanimpute.R

@@ -1,6 +1,18 @@
-#' Meanimputation
+#' Meanimpute
+#'
+#' Replace NA's with mean value
+#' @param x a numeric vector
+#' @return new vector, where NA's are replace by mean \code{x}
+#' @examples
+#' example_vector=c(1,5,NA,NA)
+#' meanimpute(example_vector)
 #' @export
 meanimpute <- function(x) {
+
+  if(is.null(x)) {stop("Input vector cannot be NULL.")}
+  if(all(is.na(x))) {stop("Input vector should contain at least one numeric element.")}
+  if(any(is.numeric(x)==FALSE)) {stop("Input vector should contain at least one numeric element.")}
+
   x[is.na(x)] <- mean(x, na.rm = TRUE)
   x
 }

R/transform_log.R

@@ -0,0 +1,20 @@
+#' transform_log
+#'
+#' log-transformation of a numeric vector. For details about log-transformation please see you basic school math textbook.
+#' @param x a numeric vector
+#' @return log-transformed vector \code{x}
+#' @examples
+#' example_vector=c(1,2,3,4,5,6,7,8,9,10)
+#' transform_log(example_vector)
+#' @export
+
+transform_log <- function(x){
+
+  if( is.null(x) )    stop("Input vector is not allowed to be NULL.")
+  if( any(is.na(x)) ) stop("There is at least one NA value in input vector.")
+  if( any(x <= 0)  )  stop("There is at least one negative value.")
+  if( any(is.numeric(x) == FALSE) )  stop("There is at least one non-numeric value.")
+  y<-log(x)
+  return(y)
+
+}

R/windsorize.R

@@ -1,10 +1,36 @@
 #' Windsorize
 #'
-#' Do some windsorization.
+#' Its purposes is to eliminate outliers in a following way. Values of (0.5 +- p/2)th quantiles are calculated and all
+#' values above(below) those quantiles are replaced by the quantiles.
+#' @param x a numeric vector
+#' @param p quantile
+#' @return Windsorized  vector \code{x}
+#' @examples
+#' example_vector=c(-1000,1,2,3,4,5,6,7,8,9,1000)
+#' windsorize(example_vector, 0.9)
+#' 
+#' example_vector=rnorm(100)
+#' windsorize(example_vector, 0.9)
 #' @export
+#' @import stats
+
 windsorize <- function(x, p = .90) {
-  q <- quantile(x, p)
-  x[x >= q] <- q
-  x
+
+  if(is.null(x)) {stop("Input vector cannot be NULL.")}
+  if(any(is.na(x))) {stop("There should be no NA's in input vector.")}
+  if(all(is.numeric(x)==FALSE)) {stop("There should only numeric values in the input vector.")}
+
+  if(is.na(p)==TRUE) {stop("Input quantile should be a number between 0 and 1 ")}
+  if(is.numeric(p)==FALSE) {stop("Input quantile should be a number between 0 and 1 ")}
+  if(p > 1) {stop("Input quantile should be a number between 0 and 1 ")}
+  if(p < 0) {stop("Input quantile should be a number between 0 and 1 ")}
+
+
+  q_u <- quantile(x, 0.5 + p/2)
+  x[x >= q_u] <- q_u
+
+  q_l <- quantile(x, 0.5 - p/2)
+  x[x <= q_l] <- q_l
+
+  return(x)
 }
-

tests/testthat.R

@@ -0,0 +1,4 @@
+library(testthat)
+library(datacleaner)
+
+test_check("datacleaner")

tests/testthat/test_correct_input_meanimpute.R

@@ -0,0 +1,7 @@
+test_that("Incorrect input of meanimpute", {
+  #tests related to input vector
+  expect_error(meanimpute(NULL),"Input vector cannot be NULL.")
+  expect_error(meanimpute(c(NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA)),"Input vector should contain at least one numeric element.")
+  expect_error(meanimpute(c(1,2,3,4,"Dracula"), "Input vector should contain at least one numeric element."))
+
+})

tests/testthat/test_correct_input_transform_log.R

@@ -0,0 +1,8 @@
+test_that("Input of transform_log() is correct.", {
+  #tests related to input vector
+  expect_error(transform_log(NULL), "Input vector is not allowed to be NULL.")
+  expect_error(transform_log(c(NA,NA,NA,6,NA,5,NA,NA,7,NA,NA)), "There is at least one NA value in input vector.")
+  expect_error(transform_log(c(1,2,3,4,5,6,7,8,"string",1000)), "There is at least one non-numeric value.")
+  expect_error(transform_log(c(1,2,3,4,5,6,7,8,-5)), "There is at least one negative value.")
+
+})

tests/testthat/test_correct_input_windsorization.R

@@ -0,0 +1,10 @@
+test_that("Incorrect input", {
+  #tests related to input vector
+  expect_error(windsorize(NULL, .9), "Input vector cannot be NULL.")
+  expect_error(windsorize(c(NA,NA,NA,NA,NA,NA,NA,NA,NA,NA,NA), .9), "There should be no NA's in input vector.")
+  expect_error(windsorize(c(-1000,1,2,3,4,5,6,7,8,"string",1000), .9), "There should only numeric values in the input vector.")
+  expect_error(windsorize(c(-1000,1,2,3,4,6,7,8,9,1000), "string"), "Input quantile should be a number between 0 and 1.")
+  expect_error(windsorize(c(-1000,1,2,3,4,6,7,8,9,1000), 2), "Input quantile should be a number between 0 and 1")
+  expect_error(windsorize(c(-1000,1,2,3,4,6,7,8,9,1000), -2), "Input quantile should be a number between 0 and 1")
+
+})

tests/testthat/test_correct_result_meanimpute.R

@@ -0,0 +1,3 @@
+test_that("NA's in vector are correctly replace by mean", {
+  expect_equal(meanimpute(c(2,4,6,NA)), c(2,4,6,4))
+})

tests/testthat/test_correct_result_transform_log.R

@@ -0,0 +1,3 @@
+test_that("Vector is correctly log-transformed", {
+  expect_equal(transform_log(c(1,1,1)), c(0,0,0))
+})

tests/testthat/test_correct_result_windsorization.R

@@ -0,0 +1,3 @@
+test_that("vector is correctly windsorized", {
+  expect_equal(windsorize(c(-1000,1,2,3,4,5,6,7,8,9,1000),0.9), c(-499.5,1,2,3,4,5,6,7,8,9,504.5) )
+})