根据其他三个列中的条件创建新列

Question

谁能帮我根据一些严格的标准创建一个新列。

数据看起来像这样

screening consistancyAPP  consistancyVAF consistancyCRF
XXX/01    Missing         50-100% target 50-100% target
XXX/03    0-49% target    0-49% target   0-49% target
XXX/08    Missing         Missing        50-100% target
XXX/13    Missing         Missing        Missing
XXX/09    Missing         0-49% target   50-100% target
XXX/04    0-49% target    0-49% target   50-100% target

df <- data.frame (screening  = c("XXX/01", "XXX/03", "XXX/08","XXX/13","XXX/09","XXX/04"),
                  consistancyAPP = c("Missing", "0-49% target", "Missing","Missing","Missing","0-49% target"),
                  consistancyVAF = c("50-100% target", "0-49% target", "Missing","Missing","0-49% target","0-49% target"),
                  consistancyCRF = c("50-100% target", "0-49% target", "50-100% target","Missing","50-100% target","50-100% target")
)

新列的条件

• 如果所有三个都相同，例如失踪然后失踪
• 如果两个缺失（例如 XXX/08），则取另一个（例如 50-100% 目标）
• 如果两个一致（例如 XXX/01），则该值（例如 50-100% 目标）
• 如果所有三个不同的 (XXX/09) 都采用较低的（0-49% 目标）

我希望输出看起来像这样

screening consistancyAPP  consistancyVAF consistancyCRF  consistancyOverall
XXX/01    Missing         50-100% target 50-100% target  50-100% target
XXX/04    0-49% target    0-49% target   0-49% target    0-49% target
XXX/08    Missing         Missing        50-100% target  50-100% target
XXX/13    Missing         Missing        Missing         Missing
XXX/09    Missing         0-49% target   50-100% target  0-49% target
XXX/04    0-49% target    0-49% target   50-100% target  0-49% target

Answer 1

好的，我终于让我的功能工作了。 对于这个问题，我创建了一个函数，该函数采用长度为 3 的字符向量，并根据上述条件返回首选选项。然后，我们使用 dplyr 函数 rowwise 和 mutate 将其应用于数据集的每一列：

score <- function(symbols) {
  
  same <- symbols[1] == symbols[2] && symbols[2] == symbols[3]
  two_same <- symbols[1] == symbols[2] || symbols[2] == symbols[[3]] ||
    symbols[1] == symbols[3]
  not_same <- symbols[1] != symbols[2] && symbols[1] != symbols[2] &&
    symbols[2] != symbols[3]
  two_missing <- sum(symbols == "Missing") == 2
  
  if(same) {
    return(symbols[1])
  } else if(not_same) {
    return("0-49% target")
  } else if(two_missing) {
    return(symbols[symbols != "Missing"])
  } else if(two_same) {
    return(symbols[duplicated(symbols)])
  }
}

x <- c("Missing", "50-100% target", "50-100% target")

score(x)
[1] "50-100% target"

y <- c("Missing", "0-49% target", "50-100% target")

score(y)
[1] "0-49% target"

z <- c("Missing", "Missing", "50-100% target")

score(z)
[1] "50-100% target"

现在我们将其应用于您的数据集，如下所示：

df %>%
  rowwise() %>%
  mutate(consistancyOverall = score(c_across(starts_with('consistancy'))))

# A tibble: 6 x 5
# Rowwise: 
  screening consistancyAPP consistancyVAF consistancyCRF consistancyOverall
  <chr>     <chr>          <chr>          <chr>          <chr>             
1 XXX/01    Missing        50-100% target 50-100% target 50-100% target    
2 XXX/03    0-49% target   0-49% target   0-49% target   0-49% target      
3 XXX/08    Missing        Missing        50-100% target 50-100% target    
4 XXX/13    Missing        Missing        Missing        Missing           
5 XXX/09    Missing        0-49% target   50-100% target 0-49% target      
6 XXX/04    0-49% target   0-49% target   50-100% target 0-49% target  

你也可以从purrr函数中使用pmap，但是，这次我们不使用rowwise，因为pmap对数据框的每一行都应用了一个函数：

df %>%
  mutate(consistancyOverall = pmap(list(consistancyAPP, consistancyVAF, consistancyCRF), 
                                   ~ score(c(...))))

  screening consistancyAPP consistancyVAF consistancyCRF consistancyOverall
1    XXX/01        Missing 50-100% target 50-100% target     50-100% target
2    XXX/03   0-49% target   0-49% target   0-49% target       0-49% target
3    XXX/08        Missing        Missing 50-100% target     50-100% target
4    XXX/13        Missing        Missing        Missing            Missing
5    XXX/09        Missing   0-49% target 50-100% target       0-49% target
6    XXX/04   0-49% target   0-49% target 50-100% target       0-49% target

Answer 2

您可以编写一个函数，根据条件返回输出并应用于每一行。

library(dplyr)

calculateOverall <- function(x) {
  tmp <- sort(table(x), decreasing = TRUE)
            #If all three are same
  case_when(n_distinct(x) == 1 ~ first(x),
            #If two missing take other
            sum(x == 'Missing') >= 2 ~ x[x!= 'Missing'][1], 
            #If two in agreement take that value
            tmp[1] >= 2 ~ names(tmp)[1], 
            #If all three different
            n_distinct(tmp) == 1 ~ '0-49% target')
}

df %>%
  rowwise() %>%
  mutate(consistancyOverall = calculateOverall(c_across(starts_with('consistancy'))))

# screening consistancyAPP consistancyVAF consistancyCRF consistancyOverall
#  <chr>     <chr>          <chr>          <chr>          <chr>             
#1 XXX/01    Missing        50-100% target 50-100% target 50-100% target    
#2 XXX/03    0-49% target   0-49% target   0-49% target   0-49% target      
#3 XXX/08    Missing        Missing        50-100% target 50-100% target    
#4 XXX/13    Missing        Missing        Missing        Missing           
#5 XXX/09    Missing        0-49% target   50-100% target 0-49% target      
#6 XXX/04    0-49% target   0-49% target   50-100% target 0-49% target      

或者purrr::pmap_chr：

df %>%
  mutate(consistancyOverall = purrr::pmap_chr(select(., starts_with('consistancy')),
                              ~calculateOverall(c(...))))