如果某个列是 NA，我如何获取匹配字符串的第一个非 NA 列

Question

I haven't seen a similar question for this issue (it's pretty specific).

I have three people who can be chosen to give answers to questions about a patient. Only two of those three people are actually used for any particular patient (in my real data it's always two who are chosen, but among a pool of 10 people).

If the initial two people disagree, a third person is used (3rdOpinion) and that opinion overrides the others.

Therefore, the final result = the 3rd opinion result, UNLESS the initial two opinions are the same (ie 3rdOpinion is NA), in which case the final result is just the opinion given by the initial two people (ie the first non-NA value for that question for that patient)

So for example patient 1 question 1, Ben and Chris disagreed, so 3rdOpinion was used as the final result.

For question 2, patient 2, Adam and Chris both said "yes", so the final result is "yes" and the 3rd opinion wasn't used.

How can I code my data to give the last two columns, Question1_final and Question2_final?

#Code to reproduce the data with the desired last two columns:
Patient <- c("1","2","3")
Question1_Adam <- c(NA,"Yes","No")
Question2_Adam <- c(NA,"Yes","No")
Question1_Ben <- c("Yes",NA,"Unlikely")
Question2_Ben <- c("No",NA,"No")
Question1_Chris <- c("Probably","Probably",NA)
Question2_Chris <- c("Unlikely","Yes",NA)
Question1_3rdOpinion <- c("Probably","Yes","No")
Question2_3rdOpinion <- c("No",NA,NA)
Question1_final <- c("Probably","Yes","No")
Question2_final <- c("No","Yes","Unlikely")
df <- data.frame(Patient, Question1_Adam, Question2_Adam, Question1_Ben, Question2_Ben, Question1_Chris, Question2_Chris, Question1_3rdOpinion, Question2_3rdOpinion, Question1_final, Question2_final)

I figured I needed something like this, but not sure how to code the last part:

df <- transform(df, Q1_final = ifelse(!is.na(Question1_3rdOpinion), Question1_3rdOpinion, *here I would somehow grep the first non-NA question 1 value*))

Answer 1

One way using dplyr and tidyr is to get data in long format, separate question and people into different columns. For each Patient and Question check if the first two values are the same take that value or take the 3rd value.

library(dplyr)
library(tidyr)

df %>%
  pivot_longer(cols = -Patient, values_drop_na = TRUE) %>%
  separate(name, into = c("Question", "name"), sep = "_") %>%
  group_by(Patient, Question) %>%
  summarise(ans = if (value[1L] == value[2L]) value[1L] else value[3L]) %>%
  mutate(Question = paste0(Question, "_final")) %>%
  pivot_wider(names_from = Question, values_from = ans) %>%
  left_join(df, by = "Patient")

# Patient Question1_final Question2_final Question1_Adam Question2_Adam
#  <fct>   <fct>           <fct>           <fct>          <fct>         
#1 1       Probably        No              NA             NA            
#2 2       Yes             Yes             Yes            Yes           
#3 3       No              No              No             No            
# … with 6 more variables: Question1_Ben <fct>, Question2_Ben <fct>,
#   Question1_Chris <fct>, Question2_Chris <fct>, Question1_3rdOpinion <fct>,
#   Question2_3rdOpinion <fct>