使用data.table [duplicate]选择组中的x最高值

Posted 2021-04-14

这个问题在这里已有答案：

• How to extract the first n rows per group? 2回答
• Select the top N values by group 7个答案

如何在data.table中为每个组选择x最高值？

例如，我想为每个组（日期）取两个最高值（Val）。所以对于这个数据集：

Date    Name    Val
01/01/2010  A   3
01/01/2010  B   2
01/01/2010  C   1
02/01/2010  A   4
02/01/2010  B   2
02/01/2010  C   3
02/01/2010  D   1

代码应该返回：

Date    Name    Val
01/01/2010  A   3
01/01/2010  B   2
02/01/2010  A   4
02/01/2010  C   3

答案

df <- read.table(text = "Date    Name    Val
01/01/2010  A   3
                 01/01/2010  B   2
                 01/01/2010  C   1
                 02/01/2010  A   4
                 02/01/2010  B   2
                 02/01/2010  C   3
                 02/01/2010  D   1", 
                 header = TRUE, stringsAsFactors = FALSE)

setDT(df)
df[, max_val := max(Val), by = Date]
df[, max_sec := order(Val, decreasing = T)[2], by = Date]
df <- df[Val == max_val | Val == max_sec, ]
df[, c("max_val", "max_sec") := NULL]

         Date Name Val
1: 01/01/2010    A   3
2: 01/01/2010    B   2
3: 02/01/2010    A   4
4: 02/01/2010    C   3