R中嵌套的if-else循环

Question

I have a data frame named "crimes" which contains a "pre_rate" column that denotes the crime rate before a certain law is implemented. I would like to put each rate in a "rate_category" column using a nested if-else loop. I have the following code:

crimes$rate_category = 
  with(crimes, ifelse(pre_rate > 0.26 && pre_rate < 0.87, 1,
    ifelse(pre_rate > 1.04 && pre_rate < 1.94, 2, 
      ifelse(pre_rate > 2.03 && pre_rate < 2.96, 3, 
        ifelse(pre_rate > 3.10 && pre_rate < 3.82, 4, 
          ifelse(pre_rate > 4.20 && pre_rate < 11.00, 5, "NA"))))))
crimes

and here's a reproducible example:

pre_rate = c(0.27, 1.91, 2.81, 3.21, 4.80) 
crimes = data.frame(pre_rate)   
crimes

However, when I run the loop with my original data frame, all levels in the "rate_category" column is incorrectly set to 1. What seems to be the problem with the loop above?

Answer 1

Instead of nesting ifelse statements might I recommend using case_when . It is a bit easier to read/follow. But as @Marius mentioned your problem is the && instead of using & .

library(tidyverse)
crimes <- data.frame(pre_rate = c(0.27, 1.91, 2.81, 3.21, 4.80))

crimes %>% 
  mutate(rate_category = case_when(pre_rate > 0.26 & pre_rate < 0.87 ~ 1,
                                   pre_rate > 1.04 & pre_rate < 1.94 ~ 2,
                                   pre_rate > 2.03 & pre_rate < 2.96 ~ 3,
                                   pre_rate > 3.10 & pre_rate < 3.82 ~ 4,
                                   pre_rate > 4.20 & pre_rate < 11.00 ~ 5))

Answer 2

Why not define your lower bounds and upper bounds in two vectors then rely on indexing? Using this method, there is no need to write pre_rate > num1 & pre_rate < num2 multiple times.

lowB <- c(0.26, 1.04, 2.03, 3.10, 4.2)
uppB <- c(0.87, 1.94, 2.96, 3.82, 11)

myCategory <- 1:5 ## this can be whatever categories you'd like

crimes$rate_category <- with(crimes, myCategory[pre_rate > lowB & pre_rate < uppB])

Answer 3

Instead of multiple nested ifelse() , a non-equi join and update on join can be used

# OP's sample data set with one out-of-bounds value appended
crimes = data.frame(pre_rate = c(0.27, 1.91, 2.81, 3.21, 4.80, 1.0))   

library(data.table)
# specify categories, lower, and upper bounds
bounds <- data.table(
  cat = 1:5,
  lower = c(0.26, 1.04, 2.03, 3.10, 4.2),
  upper = c(0.87, 1.94, 2.96, 3.82, 11)
)
# non-equi join and update on join
setDT(crimes)[bounds, on = .(pre_rate > lower, pre_rate < upper), rate_category := cat][]

  pre_rate rate_category 1: 0.27 1 2: 1.91 2 3: 2.81 3 4: 3.21 4 5: 4.80 5 6: 1.00 NA 

Note that pre-rate values which are outside of any of the given intervals do get a NA rate_category automatically.

Answer 4

You may use algebraic approach to solve your problem, it should be faster than your ifelse:

pre_rate = c(0.27, 1.91, 2.81, 3.21, 4.80) 
crimes = data.frame(pre_rate)   
crimes$rate = (pre_rate > 0.26 & pre_rate < 0.87)*1 + 
  (pre_rate > 1.04 & pre_rate < 1.94)* 2 + 
  (pre_rate > 2.03 & pre_rate < 2.96)* 3 + 
  (pre_rate > 3.10 & pre_rate < 3.82)* 4 + 
  (pre_rate > 4.20 & pre_rate < 11.00)* 5

The idea here is to just get true or false values from expression, then it is getting multiplied by the number for which you see that as a category. The only difference would be that you won't be getting NAs here for non match instead you will get a zero, which you can off course change it. Also to add, Use "&" in cases where you want to vectorize (element by element match) your outcome as mentioned in the comments.

Output:

#> crimes
# pre_rate rate
#1     0.27    1
#2     1.91    2
#3     2.81    3
#4     3.21    4
#5     4.80    5

Answer 5

If your data does not contain gaps, and you just want an index, you can use .bincode :

crimes$rate_category <- .bincode(crimes$pre_rate,
                                 breaks = c(-Inf, 1, 2, 3, 4, Inf))

If you want specific values for each interval, you can use a rolling join via the data.table package:

library(magrittr)
library(data.table)

rate_category_by_pre_rate <- 
  data.table(rate_category = c("foo", "bar", "foobar", "baz", "foobie"),
             pre_rate = c(1, 2, 3, 4, 11)) %>%
  setkey(pre_rate)

crimes %>%
  as.data.table %>%
  setkey(pre_rate) %>%
  rate_category_by_pre_rate[., roll = -Inf]

#>    rate_category pre_rate
#> 1:           foo     0.27
#> 2:           bar     1.91
#> 3:        foobar     2.81
#> 4:           baz     3.21
#> 5:        foobie     4.80

However, in your case, you may only need ceiling ( ie round-up the value of pre_rate and cap it at 5:

crimes$rate_category <- pmin(ceiling(crimes$pre_rate), 5)

#>   pre_rate rate_category
#> 1     0.27             1
#> 2     1.91             2
#> 3     2.81             3
#> 4     3.21             4
#> 5     4.80             5