mutate multiple columns based on single if condition-CodePudding

I have an example df:

df <- data.frame(
  var1 = c("A", "B"),
  var2 = c("C", "D"),
  var3 = c("E", "F"),
  var4 = c("G", "H"),
  var5 = c("I", "J"),
  var6 = c("K", "L"),
  var7 = c("M", "N"),
  var8 = c(NA, "P"))

and df2 is my desired output:


df2 <- data.frame(
  var1 = c("I", "B"),
  var2 = c("Z", "D"),
  var3 = c("M", "F"),
  var4 = c("G", "H"),
  var5 = c("I", "J"),
  var6 = c("K", "L"),
  var7 = c("M", "N"),
  var8 = c(NA, "P"),
  inputed_flag = c("Y","N"))

Basically the logic is as follows:


df3 <- df %>%
  mutate(var1 = ifelse(is.na(var8), var5, var1),
         var2 = ifelse(is.na(var8), "Z", var2),
         var3 = ifelse(is.na(var8), var7, var3),
         imputed_flag = ifelse(is.na(var8), "Y", "N"))

but in R is there an easier/more compact way of mutating columns based on a single if condition? The condition in this case is if var8 is missing, then impute certain values for other variables. I don't know another way of doing this without a bunch of ifelse statements with the same conditon.

In sas we can do something like this


if missing(var8) then do;
var1 = var5;
var2 = "Z";
var3 = var7;
imputed_flag = "Y";
end;
if not missing(var8) then imputed_flag = "N";

where if a single if condition is met, then several variables can be mutated all in one if statement.

I'm looking for an elegant R solution like this in sas if possible.

CodePudding user response：

I think simple subset assignment in base R is reasonable here:

i <- is.na(df$var8)
df[i, 1:3] <- data.frame(var1 = df$var5[i], var2 = "z", var3 = df$var7[i])
cbind(df, imputed_flag = i)
#>   var1 var2 var3 var4 var5 var6 var7 var8 imputed_flag
#> 1    I    z    M    G    I    K    M <NA>         TRUE
#> 2    B    D    F    H    J    L    N    P        FALSE

^{Created on 2022-08-17 by the reprex package (v2.0.1)}

CodePudding user response：

Seems like Map would be more concise

nm1 <- paste0("var", 1:3)
i1 <- is.na(df$var8)
df[nm1] <- Map(function(x, y) ifelse(i1, y, x), 
     df[nm1], c(df['var5'], 'Z', df['var7']))
df$imputed_flag <- c("N", "Y")[1   i1]

-checking

> names(df2)[length(df2)] <- "imputed_flag"
> all.equal(df, df2)
[1] TRUE

Or another option (inspired from @Allan Cameron's post) will be to create a temp column in 'df' and then use row index and column name reordering to update the values

nm1 <- paste0("var", 1:3)
i1 <- is.na(df$var8)
df$z <- "Z"
df[i1, nm1] <- df[i1, c("var5", "z", "var7")]
df$imputed_flag <- c("N", "Y")[1   i1]
df$z <- NULL

With across2 (from dplyover), we could do

library(dplyover)
library(dplyr)
library(stringr)
df %>% 
  mutate(z = 'Z',
   across2(var1:var3, c(var5, z, var7),  
   ~ case_when(is.na(var8)~ .y, TRUE ~ .x), 
   .names_fn = ~ str_remove(.x, "_.*")), 
   imputed_flag = c("N", "Y")[1   is.na(var8)])
  var1 var2 var3 var4 var5 var6 var7 var8 z imputed_flag
1    I    Z    M    G    I    K    M <NA> Z            Y
2    B    D    F    H    J    L    N    P Z            N