R从数据框中的序数中选择第一项

Question

Suppose I have a dataframe:

df=data.frame(cat=c("b1","b2","b3","b2","b5","b1","b3"),
              item=c("a1","a2","a3","a4","a1","a3","a4"),
              status=c("ok","good","bad","excellent","ok","good","bad"))

And I need for each category b1-b5, choose only the top a (ranked by status from excellent to good to ok to bad) and corresponding status, and in case of tie take a random one.

So for b1 it'll take a3 good instead of a1 ok, for b3 it could take either a3 bad or a4 bad. sample output:

  cat item    status
  b1   a3     good

What's the best way to do this?

Answer 1

You could try:

  df$status <- factor(df$status, levels=c("excellent", "good", "ok", "bad"))
  library(dplyr)
  df %>%
  group_by(cat) %>% 
  arrange(status)  %>% 
  filter(row_number()==1)
  #  cat item    status
  #1  b2   a4 excellent
  #2  b1   a3      good
  #3  b5   a1        ok
  #4  b3   a3       bad

Or using data.table

 library(data.table)
 setDT(df)[,.SD[order(status)][1], by=cat]
 #   cat item    status
 #1:  b1   a3      good
 #2:  b2   a4 excellent
 #3:  b3   a3       bad
 #4:  b5   a1        ok

Update

I noticed that you wanted to get a random sample in cases of tie

 setDT(df)[, if(length(status)>1 & length(unique(status))==1) 
               .SD[sample(1:.N,1)]
                 else .SD[order(status)][1]
                                     , by=cat]
  # cat item    status
 #1:  b1   a3      good
 #2:  b2   a4 excellent
 #3:  b3   a3       bad
 #4:  b5   a1        ok


  df[, if(length(status)>1 & length(unique(status))==1)
             .SD[sample(1:.N,1)]
               else .SD[order(status)][1] , 
                                    by=cat]
#   cat item    status
#1:  b1   a3      good
#2:  b2   a4 excellent
#3:  b3   a4       bad
#4:  b5   a1        ok

Answer 2

First put status in correct order:

R>df$status <- ordered(df$status, c("bad", "ok", "good", "excellent"))

Then :

R>by(df, df$cat, function(d) d[which.max(d$status), ])
df$cat: b1
  cat item status
6  b1   a3   good
------------------------------------------------------------ 
df$cat: b2
  cat item    status
4  b2   a4 excellent
------------------------------------------------------------ 
df$cat: b3
  cat item status
3  b3   a3    bad
------------------------------------------------------------ 
df$cat: b5
  cat item status
5  b5   a1     ok