R(統計) 変数毎に基本統計量(summary)を算出する。int,num,factorに対応。

基礎集計や機械学習を実行する前に、異常値や基礎情報を知りたいために作成しました。

integerやnumericだけでなく、factorやlogicalにも対応しています。

まだエラーや書き方がRらしくないかもしれませんが、コメント頂けると幸いです。

先に結果から。

> output <- mySummary(iris)
> output
      colnames   class                                 inform   sum     sd   mean  Min Median  Max NA.s
2 Sepal.Length numeric                                   <NA> 876.5 0.8281 5.8433  4.3    5.8  7.9    0
3  Sepal.Width numeric                                   <NA> 458.6 0.4359 3.0573    2      3  4.4    0
4 Petal.Length numeric                                   <NA> 563.7 1.7653  3.758    1   4.35  6.9    0
5  Petal.Width numeric                                   <NA> 179.9 0.7622 1.1993  0.1    1.3  2.5    0
6      Species  factor setosa:50, versicolor:50, virginica:50  <NA>   <NA>   <NA> <NA>   <NA> <NA>    0

プログラム

# 基本統計量を算出
mySummary <- function(df){
  # only data frame
  if (!is.data.frame(df)) return(NA)
  # create new df
  new.df <- data.frame(
    colnames  = NA
    ,class     = NA
    ,inform    = NA
    ,sum       = NA
    ,sd        = NA
    ,mean      = NA
    ,Min       = NA
    ,Median    = NA
    ,Max       = NA
    ,"NA's"    = NA
  )
  col.count <- ncol(df)
  for (i in 1:col.count) {
    if (is.integer(df[,i]) || is.numeric(df[,i])) {
      cal.sum    <- sum(as.numeric(df[,i]), na.rm=TRUE)
      cal.mean   <- mean(df[,i], na.rm=TRUE)
      cal.sd     <- sd(df[,i], na.rm=TRUE)
      cal.min    <- min(df[,i], na.rm=TRUE)
      cal.median <- median(df[,i], na.rm=TRUE)
      cal.max    <- max(df[,i], na.rm=TRUE)
    } else if (is.factor(df[,i]) || is.logical(df[,i])) {
      tmp <- summary(df[,i])
      str.val <- paste(names(tmp), tmp, sep=":")
      cal.summary<- paste(str.val, collapse=", ")
    }

    insert.row <- c(
      colnames  = colnames(df[i])
      ,class     = class(df[,i])
      ,inform    = ifelse(exists("cal.summary"), cal.summary, NA)
      ,sum       = ifelse(exists("cal.sum"), cal.sum, NA)
      ,sd        = ifelse(exists("cal.sd"), round(cal.sd, 4), NA)
      ,mean      = ifelse(exists("cal.mean"), round(cal.mean, 4), NA)
      ,Min       = ifelse(exists("cal.min"), cal.min, NA)
      ,Median    = ifelse(exists("cal.median"), cal.median, NA)
      ,Max       = ifelse(exists("cal.max"), cal.max, NA)
      ,"NA's"    = sum(is.na(df[,i]))
    )
    new.df <- rbind(new.df, insert.row)
    suppressWarnings(rm("cal.sum","cal.mean","cal.max","cal.median","cal.min","cal.sd","cal.summary"))
  }
  new.df <- new.df[-1,]
  return(new.df)
}