CODEWITHSUNDEEP
rif-statementtidyversedplyr
Ian.T
Ian.T

Reputation: 1166

Mutate_at and ifelse problem with vectorization

I thinks that is easier if i show, what is the problem, So I have this numeric data

  MoSold YrSold SalePrice OverallQual OverallCond
1      2      3    208500           7           5
2      5      2    181500           6           8
3      9      3    223500           7           5
4      2      1    140000           7           5
5     12      3    250000           8           5
6     10      4    143000           5           5

thanks to mutate_at and a nested ifelse i would like to change each row if condition is true (the column mean is higher than 0.8) however when i try to do it whith this code

data %>% 
  mutate_at(vars(MoSold, YrSold, SalePrice, OverallQual, OverallCond), 
            ~(ifelse((mean(., na.rm = T)) > 4, log(.), .))) %>% head()

I get the following data, all the columns have the same value

     MoSold YrSold SalePrice OverallQual OverallCond
1 0.6931472      3  12.24769     1.94591    1.609438
2 0.6931472      3  12.24769     1.94591    1.609438
3 0.6931472      3  12.24769     1.94591    1.609438
4 0.6931472      3  12.24769     1.94591    1.609438
5 0.6931472      3  12.24769     1.94591    1.609438
6 0.6931472      3  12.24769     1.94591    1.609438

and i would like to have the log of the corresponing value for each row if the condition is true and the raw value if condition is false

I know that one solution is to use a for loop but a would really like a solution with dplyr/tidyverse

Thanks in advance

I.

Upvotes: 4

Views: 3298

Answers (1)

akrun
akrun

Reputation: 887088

The issue is related to mean used as the test for ifelse, which is a single value while the 'yes', 'no' arguments are of different length, i.e. the logical expression with that result in a single TRUE/FALSE and this gets replicated for the full length with recycling the first element of 'yes', 'no'

Here, we can use if/else instead of ifelse

library(dplyr)
data %>%
   mutate_all(~ if(mean(., na.rm = TRUE) > 4) log(.) else .)

In the dplyr 1.0.0, an option is mutate/across

data %>%
   mutate(across(everything(),
         ~ if(mean(., na.rm = TRUE) > 4) log(.) else .))
#  MoSold YrSold SalePrice OverallQual OverallCond
#1 0.6931472      3  12.24769    1.945910    1.609438
#2 1.6094379      2  12.10901    1.791759    2.079442
#3 2.1972246      3  12.31717    1.945910    1.609438
#4 0.6931472      1  11.84940    1.945910    1.609438
#5 2.4849066      3  12.42922    2.079442    1.609438
#6 2.3025851      4  11.87060    1.609438    1.609438

If we want to use ifelse, replicate the single logical value to make all the 'test', 'yes', 'no' of same length

data %>% 
   mutate_at(vars(MoSold, YrSold, SalePrice, OverallQual, OverallCond), 
             ~(ifelse(rep((mean(., na.rm = T)) > 4, n()), log(.), .)))
#     MoSold YrSold SalePrice OverallQual OverallCond
#1 0.6931472      3  12.24769    1.945910    1.609438
#2 1.6094379      2  12.10901    1.791759    2.079442
#3 2.1972246      3  12.31717    1.945910    1.609438
#4 0.6931472      1  11.84940    1.945910    1.609438
#5 2.4849066      3  12.42922    2.079442    1.609438
#6 2.3025851      4  11.87060    1.609438    1.609438

data

data <- structure(list(MoSold = c(2L, 5L, 9L, 2L, 12L, 10L), YrSold = c(3L, 
2L, 3L, 1L, 3L, 4L), SalePrice = c(208500L, 181500L, 223500L, 
140000L, 250000L, 143000L), OverallQual = c(7L, 6L, 7L, 7L, 8L, 
5L), OverallCond = c(5L, 8L, 5L, 5L, 5L, 5L)), class = "data.frame",
row.names = c("1", 
"2", "3", "4", "5", "6"))

Upvotes: 5

Related Questions