Calculate the rowwise mean when a maximum number of NA values is given for a set of columns using dp-CodePudding

Example dataset...

> tribble(
    ~colA, ~colB, ~colC, ~colD, ~colE,
    1, 2, 3, 4, 5,
    2, 3, NA, 4, 5,
    3, NA, NA, NA, 4,
    4, NA, NA, 5, 6
  )
# A tibble: 4 × 5
   colA  colB  colC  colD  colE
  <dbl> <dbl> <dbl> <dbl> <dbl>
1     1     2     3     4     5
2     2     3    NA     4     5
3     3    NA    NA    NA     4
4     4    NA    NA     5     6

How can I create a new column giving the mean of columns B, C, D and E if only two (at most) NAs are present? In this case, the third row mean should be NA as it has 3 NAs. I have put colA because I want to be able to use tidyselect to choose which variables are included.

So far I have this...

dat %>% 
  rowwise() %>% 
  mutate(
    mean = if_else(
      c_across(colB, colC, colD, colE), 
      condition = sum(is.na(.)) <= 2, 
      true = mean(., na.rm = T), 
      false = NA
      )
    )

But I get this error message...

Error in `mutate()`:
! Problem while computing `mean = if_else(...)`.
ℹ The error occurred in row 1.
Caused by error in `if_else()`:
! `false` must be a double vector, not a logical vector.
Run `rlang::last_error()` to see where the error occurred.
Warning message:
Problem while computing `mean = if_else(...)`.
ℹ argument is not numeric or logical: returning NA
ℹ The warning occurred in row 1.

In an ideal world, I would have a function for taking the rowwise mean for a set of columns and a given number of allowed NAs that I could repurpose.

CodePudding user response：

We can use across to select column of interest.

library(dplyr)

dat %>% 
  mutate(mean = ifelse(rowSums(is.na(across(-colA))) > 2, 
                       NA, 
                       rowMeans(across(-colA), na.rm = T)))

# A tibble: 4 × 6
   colA  colB  colC  colD  colE  mean
  <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
1     1     2     3     4     5   3.5
2     2     3    NA     4     5   4  
3     3    NA    NA    NA     4  NA  
4     4    NA    NA     5     6   5.5

CodePudding user response：

In base R:

df$mean <- apply(df[-1], 1, \(x) if (sum(is.na(x)) <= 2) mean(x, na.rm = T) else NA)

df

#> # A tibble: 4 x 6
#>    colA  colB  colC  colD  colE  mean
#>   <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1     1     2     3     4     5   3.5
#> 2     2     3    NA     4     5   4  
#> 3     3    NA    NA    NA     4  NA  
#> 4     4    NA    NA     5     6   5.5

Or using dplyr:

library(dplyr)

df %>% 
  mutate(mean = apply(.[-1], 1, \(x) if (sum(is.na(x)) <= 2) mean(x, na.rm = T) else NA))

#> # A tibble: 4 x 6
#>    colA  colB  colC  colD  colE  mean
#>   <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1     1     2     3     4     5   3.5
#> 2     2     3    NA     4     5   4  
#> 3     3    NA    NA    NA     4  NA  
#> 4     4    NA    NA     5     6   5.5

CodePudding user response：

We can do the following. This is an example how to select a set of columns with select in rowSums and rowMeans.

library(dplyr)

dat2 <- dat %>%
  mutate(mean = ifelse(rowSums(is.na(select(., -colA))) > 2, 
                       NA, 
                       rowMeans(select(., -colA), na.rm = TRUE)))

CodePudding user response：

data.table option:

library(data.table)
setDT(df)[!rowSums(is.na(df)) > 2, mean := rowMeans(.SD, na.rm = TRUE), .SDcols = -1]

Output:

   colA colB colC colD colE mean
1:    1    2    3    4    5  3.5
2:    2    3   NA    4    5  4.0
3:    3   NA   NA   NA    4   NA
4:    4   NA   NA    5    6  5.5