R-获取汇总表，该汇总表包含数据框的指定百分比级别

问题描述

我想获得一个汇总表，该表显示的内容比R中的summary(x)函数生成的典型描述统计数据还要多。例如10％的百分位数，90％的百分位数。 我在网上找到的其他答案都推荐了给出答案的方法，但不是列表形式。

我一直在寻找一种方法，可以仅在summary（x）函数生成的摘要表中添加指定的百分位数级别。

以下是示例数据：

df = data.frame("a"=seq(1,10),"b"=seq(10,100,"c"=letters[seq(1,10)],"d"=seq(5,95,10))

解决方法

# generate data
df = data.frame("a"=seq(1,10),"b"=seq(10,100,"c"=letters[seq(1,10)],"d"=seq(5,95,10))

# filter numerical columns
ndf = Filter(is.numeric,df)
features = colnames(ndf)

# percentiles reqd
p_reqd = c(0,0.10,0.25,0.5,0.75,0.90,0.95,1)   # more percentile levels can be specified here
                                               # after adding/removing,adjust p_lev as well

# labels for specified percentiles + mean
p_lev = c('Min','10%','25%','50%','Mean','75%','90%','95%','Max')

# created empty dataframe with row names specified
final = data.frame(row.names = p_lev)

# loop
for (i in features) {
  x = ndf[,i]
  sm = data.frame("dStats" = quantile(x,p_reqd))
  final[1:which(rownames(final)=="50%"),i] = sm$dStats[1:which(rownames(sm)=="50%")]
  final[which(rownames(final)=="50%")+1,i] = round(mean(x),2)
  final[(which(rownames(final)=="50%")+2):nrow(final),i] = 
    sm$dStats[(which(rownames(sm)=="50%")+1):nrow(sm)]  
}

# custom summary table
final

,

还有一种dplyr和tidyr的方式。

df = data.frame("a"=seq(1,10))
library(dplyr)
library(tidyr)
out <- df %>% summarise_if(is.numeric,.funs = list(
  "Min" = min,"10%" = function(x)quantile(x,.1),"25%" = function(x)quantile(x,.25),"50%" = median,"Mean" = mean,"75%" = function(x)quantile(x,.75),"90%" = function(x)quantile(x,.90),"Max" = max)) %>% 
pivot_longer(cols=everything(),names_pattern = "(.*)_(.*)",names_to = c("var","stat"),values_to="vals") %>% 
  pivot_wider(names_from="var",values_from="vals",id_cols="stat") %>% 
  as.data.frame()

rownames(out) <- out$stat
out <- out %>% select(-stat)
out
#          a     b    d
# Min   1.00  10.0  5.0
# 10%   1.90  19.0 14.0
# 25%   3.25  32.5 27.5
# 50%   5.50  55.0 50.0
# Mean  5.50  55.0 50.0
# 75%   7.75  77.5 72.5
# 90%   9.10  91.0 86.0
# Max  10.00 100.0 95.0