如何将 TukeyHSD 的结果排列到表格中？

Question

我有下面 TukeyHSD 的 output（变成数据框）：

df <- read.table(header=TRUE, stringsAsFactors=FALSE, text="
comp       diff         lwr        upr     p_adj
duration:5-duration:5              2.125000e-01 -0.13653578  0.5615358 0.4873403
speed:5-probability:5                 2.250000e-01 -0.12403578  0.5740358 0.4219408
probability:10-probability:5          3.875000e-01  0.03846422  0.7365358 0.0206341
duration:10-duration:5             6.875000e-01  0.33846422  1.0365358 0.0000020
speed:10-probability:5                2.250000e-01 -0.12403578  0.5740358 0.4219408
probability:60-probability:5          1.250000e-02 -0.31064434  0.3356443 0.9999974
probability:10-probability:60          1.250000e-02 -0.31064434  0.3356443 0.9999974
probability:10-speed:5          1.250000e-02 -0.31064434  0.3356443 0.9999974
                 ")

我想要下面的东西：

               duration5 probability5 speed5 duration10 probability10 speed10
duration5 
probability5   p         
speed5         p         p
duration10     p         p            p
probability10  p         p            p      p  
speed10        p         p            p      p          p

我已经尝试过这里提出的类似解决方案。

我更改了代码以识别连字符“-”上的拆分，但它不起作用（见下文）。 为什么不执行？ 有替代方法吗？

transformTable <- function(tbl, metric) {
  # Takes table of TurkeyHSD output metrics
  # and transforms them into a pairwise comparison matrix.
  # tbl is assumed to be a data.frame or tibble,
  # var is assumed to be a character string
  # giving the variable name of the metric in question
  # (here: "diff", "lwr", "upr", or "p_adj")
  tbl <- tbl %>%
    # Split comparison into individual variables
    mutate(
      Var1 = sub("\\-.*", "", comp), #before hypen
      Var2 = sub(".*-", "", comp)) # after hyphen%>%
    # Only keep relevant fields
    select(Var1, Var2, matches(metric)) %>%
    # Filter out NA's
    filter(!is.na(metric)) %>%
    # Make into "wide" format using Va r2
    spread_(key = 'Var2', value = metric, fill = '')
  
  # Let's change the row names to Var1
  row.names(tbl) <- tbl$Var1
  # And drop the Var1 column
  tbl <- select(tbl, -Var1)
  
  return(tbl)
}

transformTable(df,'p_adj')

Answer 1

一种快速而肮脏的方法

library(tidyr)
library(stringr)


df <- df %>%
  separate(col = comp, into = c('x', 'y'),  sep = '-') %>%
  mutate(x = str_remove(x, ":")) %>%
  mutate(y = str_remove(y, ":")) %>%
  select(x, y, p_adj)

df1 <- data.frame(matrix(nrow = length(unique(c(df$x, df$y))), ncol = length(unique(c(df$x, df$y)))))
colnames(df1) <- unique(c(df$x, df$y))
rownames(df1) <- unique(c(df$x, df$y))

for(i in 1:length(unique(c(df$x, df$y)))){
  for(j in 1:length(unique(c(df$x, df$y)))){
    value <- (df %>% filter(x == rownames(df1)[i]) %>% filter(y == colnames(df1)[j]) %>% select(p_adj))$p_adj
    if(length(value) != 0){
      df1[i,j] <-value
      df1[j,i] <- value
    } 
  }
}