获取与另一个变量的汇总计数的唯一值组合

Question

我有一个数据帧，看起来像这样：

pred1 pred2 pred3 exp
a     b     c     0
a     d     c     0
a     b     c     1

我想要做的是首先获得pred1-3的所有唯一组合，将它们写到一个附加表中，为每个组合的频率添加一列，并为每个组合添加另一列，该列给出了exp的值1的比例(只能是0或1)。如下所示：

pred1 pred2 pred3 freq exp_prop
a     b     c     2    0.5
a     d     c     1    0

事实证明，使用plyr的前三步非常简单：

ddply(df, .(pred1, pred2, pred3), summarise, freq=length(exp))

或更短

count(df[,c(pred1, pred2, pred3)])

但我就是想不出怎么才能得到exp的比例。

Answer 1

您就快完成了。只需在ddply命令中添加exp_prop = mean(exp)：

ddply(df, .(pred1, pred2, pred3), summarise,
      freq = length(exp), exp_prop = mean(exp))

  pred1 pred2 pred3 freq exp_prop
1     a     b     c    2      0.5
2     a     d     c    1      0.0

Answer 2

# read in your data
x <- 
read.table(text="pred1 pred2 pred3 exp
a     b     c     0
a     d     c     0
a     b     c     1" , h = T)

library(sqldf)
sqldf( "select pred1, pred2, pred3, count(*) as numtimes, avg( exp ) as prop from x group by pred1, pred2, pred3" )

###### alternative:

# write all the column names according to some pattern
cols <- paste0("pred" , 1:3 , collapse = "," )

# save your data frame to another object
y <-
    sqldf( 
        paste( 
            "select" , 
            cols  , 
            " , count(*) as numtimes, avg( exp ) as prop from x group by" , 
            cols 
        ) 
    )

# print to screen
y