[SOLVED] How to impute a conditional row-wise imputation of a constant

How to impute a conditional row-wise imputation of a constant

I am somewhat of an R newbie, am struggling with writing code for what seems like simple logic, and would appreciate any help! I am trying to impute a constant value of 1 for NA cells in each row of my data set but only for rows that have 2 or less NA cells. Ultimately, I will also be computing a new column with row-wise means after imputation. If one line of code code automagically achieve all of these things, that would be great!

Here is an example data set to work with.

tData <- data.frame(subID=c(1001,1002,1003,1004),
b1=c(1,1,2,NA),
b2=c(NA,1,1,NA),
b3=c(NA,2,2,NA),
b4=c(2,NA,1,NA))

I have been looking at various base and dplyr code examples but am riding the struggle bus.

Solution

We can do this like this:

library(dplyr)

tData %>% 
  mutate(across(-subID, ~ifelse(rowSums(is.na(tData[2:5])) <= 2 & is.na(.), 1, .))) %>%
  rowwise() %>%
  mutate(mean_value = mean(c_across(-subID), na.rm = TRUE))

 subID    b1    b2    b3    b4 mean_value
  <dbl> <dbl> <dbl> <dbl> <dbl>      <dbl>
1  1001     1     1     1     2       1.25
2  1002     1     1     2     1       1.25
3  1003     2     1     2     1       1.5 
4  1004    NA    NA    NA    NA     NaN