Replace strings with values across multiple columns at once

Question

I need to replace strings with numbers across multiple columns. Below is a sample data set:

x <- c("Low Outlier", "High Outlier", "Novice", "Novice", "Emerging", NA, "Proficient", "Approaching")
y <- c("Novice", "Approaching", "Proficient", "Approaching", "High Outlier", "Proficient",NA, "Emerging")
z <- c("High Outlier", "Proficient", "Approaching", "Emerging", "Low Outlier", "Approaching", "Approaching", "Emerging")

sam <- cbind(x,y,z)

I need to convert the "High/Low Outliers" to 0, The NA's to be left as NA, "Novice" to 1, "Emerging" to 2, "Approaching to 3, and "Proficient" to 4.

I have tried to convert a single variable with

sam$x.r <- recode(sam$x.r,'Low Outlier'=0,'High Outlier'=0,'Novice'=1,'Emerging'=2,'Approaching'=3, 'Proficient'=4)

I received an error message of "Warning message: In recode.numeric(Dat17_18.1$I.E.ScoreStat, Low Outlier = 0, High Outlier = 0, : NAs introduced by coercion"

I am not sure how to recode all of the variables at once.

score 4 · Answer 1 · answered Apr 11 '19 at 14:07

Simply do this-

sam[] <- recode(sam,'Low Outlier'=0,
                    'High Outlier'=0,
                    'Novice'=1,
                    'Emerging'=2,
                    'Approaching'=3, 
                    'Proficient'=4)

> sam
     x   y   z  
[1,] "0" "1" "0"
[2,] "0" "3" "4"
[3,] "1" "4" "3"
[4,] "1" "3" "2"
[5,] "2" "0" "0"
[6,] NA  "4" "3"
[7,] "4" NA  "3"
[8,] "3" "2" "2"

score 2 · Answer 2 · answered Apr 11 '19 at 14:01

We can use case_when from dplyr for such cases

library(dplyr)

sam %>%
   mutate_all(~case_when(. %in% c("Low Outlier", "High Outlier") ~ '0', 
                   . == "Novice" ~ '1', 
                   . == "Emerging" ~ '2', 
                   . == "Approaching" ~ '3', 
                   . == "Proficient" ~ '4', 
                   TRUE ~ NA_character_))


#     x    y z
#1    0    1 0
#2    0    3 4
#3    1    4 3
#4    1    3 2
#5    2    0 0
#6 <NA>    4 3
#7    4 <NA> 3
#8    3    2 2

However, the final output has character columns since our original columns were characters as well. We can add mutate_all(as.numeric) to convert them to numeric if needed.

data

x <- c("Low Outlier", "High Outlier", "Novice", "Novice", "Emerging", NA, 
      "Proficient", "Approaching")
y <- c("Novice", "Approaching", "Proficient", "Approaching", "High Outlier", 
      "Proficient",NA, "Emerging")
z <- c("High Outlier", "Proficient", "Approaching", "Emerging", "Low Outlier", 
      "Approaching", "Approaching", "Emerging")
sam <- data.frame(x,y,z, stringsAsFactors = FALSE)

NelsonGon · Accepted Answer · 2019-04-11T14:09:05.613

Got really repetitive real quick. Here's a simple function:

my_replacer<-function(df,y,z){    
df<-as.data.frame(apply(df,2,function(x) gsub(y,z,x)))
    #y is what you want to replace
    #z is the replacement
    #This uses regex
      df
    }
    my_replacer(sam,"Emerging.*","2")

Here is how I've used it:

library(dplyr)#can use ifelse. Still repetitive

    sam<-as.data.frame(sam)

    sam %>% 
      mutate_if(is.factor,as.character)->sam
    my_replacer(sam,"Emerging.*","2")

Result:

               x            y            z
    1  Low Outlier       Novice High Outlier
    2 High Outlier  Approaching   Proficient
    3       Novice   Proficient  Approaching
    4       Novice  Approaching            2
    5            2 High Outlier  Low Outlier
    6         <NA>   Proficient  Approaching
    7   Proficient         <NA>  Approaching
    8  Approaching            2            2

Replace others:

my_replacer(sam,"Novi.*","1")
             x            y            z
1  Low Outlier            1 High Outlier
2 High Outlier  Approaching   Proficient
3            1   Proficient  Approaching
4            1  Approaching     Emerging
5     Emerging High Outlier  Low Outlier
6         <NA>   Proficient  Approaching
7   Proficient         <NA>  Approaching
8  Approaching     Emerging     Emerging

score 2 · Answer 4 · answered Apr 11 '19 at 14:08

2

I would use named vectors as a mapping

library(dplyr)
mapping = c("High Outlier" = 0, "Low Outlier" = 0, "Novice" = 1, "Emerging" = 2, "Approaching" = 3, "Proficient" = 4)

sam %>% 
  as.data.frame() %>% 
  mutate_all(function(i) mapping[i])

answered Apr 11 '19 at 14:08

DS_UNI

2,600
2
11
22

score 0 · Answer 5 · answered Apr 11 '19 at 14:28

Another solution using factors to recode, and approxfun to assign values:

sam[] <- approxfun(1:5, c(0:3, 0))(
      as.numeric(factor(sam, 
                        c("Low Outlier", "Novice",  
                          "Emerging", "Approaching", 
                          "Proficient", "High Outlier"))))

#      x   y   z  
# [1,] "0" "1" NA 
# [2,] NA  "3" "0"
# [3,] "1" "0" "3"
# [4,] "1" "3" "2"
# [5,] "2" NA  "0"
# [6,] NA  "0" "3"
# [7,] "0" NA  "3"
# [8,] "3" "2" "2"

Replace strings with values across multiple columns at once

5 Answers5