Update of dataclean

NAMESPACE

@@ -1 +1,2 @@
 exportPattern("^[[:alpha:]]+")
+importFrom("stats", "quantile")

R/meanimpute.R

@@ -1,4 +1,6 @@
 #' Meanimputation
+#' Calculates mean of a given vector, ignores NA values.
+#' @param x A vector
 #' @export
 meanimpute <- function(x) {
   x[is.na(x)] <- mean(x, na.rm = TRUE)

R/transform_log.R

@@ -0,0 +1,18 @@
+#' transform_log 
+#' Transform numerical values into their log values
+#' @param x A vector
+#' @return logarithm of x
+#' @examples
+#' transform_log(c(NA,0,-1,exp(2)))
+#'@export 
+transform_log<-function(x){
+  if(!is.numeric(x))stop("function is expecting only numeric values")
+  x_nan<-is.na(x)
+  x[x_nan]<-1
+  ifelse(x<0,"OK", warning("input vector contains negative values, turned into NA"))
+  y<-log(x[x>=0])
+  x[x>=0]<-y
+  x[x<0]<-NA
+  x[x_nan]<-NA
+  x
+}

R/windsorize.R

@@ -1,10 +1,23 @@
 #' Windsorize
 #'
-#' Do some windsorization.
+#'
+#' Transform all outliner data to
+#' (1-p)/2 percentile value for lower outliers and 
+#' (1+p)/2 for higher outliers.
+#' 
+#' @param x A vector.
+#' @param p A quantile.
+#' @return inuput vector x with trimmed outliers by (1-p) percentile.
+#' @examples 
+#' windsorize(rnorm(100,0,1))
 #' @export
 windsorize <- function(x, p = .90) {
-  q <- quantile(x, p)
-  x[x >= q] <- q
+  if(is.null(x))stop("vector is empty")
+  y<-x[!is.na(x)]
+  if(length(y)==0)stop("vector contains only NAs")
+  q_max <- quantile(y, (1+p)/2)
+  q_min<- quantile(y,(1-p)/2)
+  x[x >= q_max] <- q_max
+  x[x<= q_min]<-q_min
   x
-}
-
+}