Creating new column based on character values in R-CodePudding

I have a data frame with a column called ‘full_name’ that presents two teams, for example: • ‘Man U to win Liverpool to win’ • ‘Liverpool to win Man U to win’ • ‘Chelsea to win Arsenal to win’ And so on…

I would like to be able to differentiate the teams into North and South, so that if ‘Man U to win Liverpool to win’ or ‘Liverpool to win Man U to win’ are presented, then this is coded as ‘North’, whereas if ‘Chelsea to win Arsenal to win’ is presented, this is coded as ‘South’, and so on.

levels(raw_data$full_name)[levels(raw_data$full_name)== "Man U to win Liverpool to win"] <- 'North'
levels(raw_data$full_name)[levels(raw_data$full_name)== "Liverpool to win Man U to win"] <- 'North'
levels(raw_data$full_name)[levels(raw_data$full_name)== "Chelsea to win Arsenal to win"] <- 'South'

The code above does not produce any error, however the dataframe remains unchanged, and there is not producing the desired output. Is a way to do this?

CodePudding user response：

Here is an alternative approach:

library(dplyr)
library(stringr)

north <- c("Liverpool|Man")
south <- c("Chelsea|Arsenal")

df %>% 
  mutate(region = case_when(str_detect(full_name, north) ~ "North",
                            str_detect(full_name, south) ~ "South",
                            TRUE ~ NA_character_))

                      full_name region
1 Liverpool to win Man U to win  North
2 Chelsea to win Arsenal to win  South
3 Man U to win Liverpool to win  North
4 Chelsea to win Arsenal to win  South
5 Liverpool to win Man U to win  North

CodePudding user response：

Here an example with a tidyverse approach that might help you

library(dplyr)

north <- c("Man U to win Liverpool to win","Liverpool to win Man U to win")
south <- c("Chelsea to win Arsenal to win")


df <- 
  data.frame(full_name = sample(c(north,south),size = 5,replace = TRUE))
             
df %>% 
  mutate(region = case_when(
    full_name %in% north ~ "North",
    full_name %in% south ~ "South"
  ))

                      full_name region
1 Chelsea to win Arsenal to win  South
2 Man U to win Liverpool to win  North
3 Chelsea to win Arsenal to win  South
4 Man U to win Liverpool to win  North
5 Man U to win Liverpool to win  North

CodePudding user response：

In base R, your code will work as intended if you remove the levels() calls. You can call factor() after replacing values if you want the column to be a factor.

# example data
raw_data <- data.frame(full_name = c(
  "Man U to win Liverpool to win", 
  "Liverpool to win Man U to win",
  "Chelsea to win Arsenal to win"
))

raw_data$full_name[raw_data$full_name == "Man U to win Liverpool to win"] <- "North"
raw_data$full_name[raw_data$full_name == "Liverpool to win Man U to win"] <- "North"
raw_data$full_name[raw_data$full_name == "Chelsea to win Arsenal to win"] <- "South"

raw_data$full_name <- factor(raw_data$full_name)

Alternatively, you can use a named vector as a lookup table:

lookup <- c(
  "Man U to win Liverpool to win" = "North",
  "Liverpool to win Man U to win" = "North",
  "Chelsea to win Arsenal to win" = "South"
)

raw_data$full_name <- factor(lookup[raw_data$full_name])

Result from either approach:

#> raw_data
  full_name
1     North
2     North
3     South

#> levels(raw_data$full_name)
[1] "North" "South"

CodePudding user response：

Here is an option with fct_recode

library(forcats)
raw_data$full_name <- with(raw_data, fct_recode(full_name, 
   North =  "Man U to win Liverpool to win",
   North = "Liverpool to win Man U to win",
   South  =  "Chelsea to win Arsenal to win"))

Or using base R

factor(raw_data$full_name, levels = c("Chelsea to win Arsenal to win", 
"Liverpool to win Man U to win", "Man U to win Liverpool to win"
), labels = c("South", "North", "North"))

Or if we want to use levels

lvls_to_change <-  c("Man U to win Liverpool to win",
   "Liverpool to win Man U to win", "Chelsea to win Arsenal to win")
lvsl_new <- c("North", "North", "South")
i1 <- levels(raw_data$full_name) %in% lvls_to_change
levels(raw_data$full_name)[i1] <- lvsl_new[match(levels(raw_data$full_name)[i1], lvls_to_change)]

data

raw_data <- structure(list(full_name = structure(c(2L, 2L, 3L, 2L,
 1L), levels = c("Chelsea to win Arsenal to win", 
"Liverpool to win Man U to win", "Man U to win Liverpool to win"
), class = "factor")), row.names = c(NA, -5L), class = "data.frame")