计算数据框中值之间交互的比例

Question

我有一个数据框，它的结构与这个玩具数据集相似。与其处理二进制数据，我宁愿生成新的列来表示二进制数据列（value1、value2 和 value3）之间的值的交互，因为只有 8 种可能的值组合（例如，TRUE.TRUE.TRUE , TRUE.TRUE.FALSE, 等等).

具体来说，我想计算每个对象和主题的每个组合的比例。

subject     object      value1     value2   value3 

1           A           TRUE       TRUE     FALSE
1           A           TRUE       TRUE     TRUE
1           B           TRUE       FALSE    TRUE
1           B           TRUE       FALSE    TRUE
1           B           TRUE       TRUE     TRUE
2           B           TRUE       FALSE    FALSE
2           A           TRUE       TRUE     FALSE
2           B           FALSE      FALSE    FALSE
3           A           TRUE       TRUE     FALSE
3           B           FALSE      TRUE     FALSE
3           B           TRUE       TRUE     TRUE
...         ...         ...        ...      ...

期望的输出：

subject     object      combination        value    
1           A           True.True.True    .5                 
1           A           True.True.False   .5   
1           B           True.True.True    .33
1           B           True.False.True   .67              
...
etc for subject 2 and 3...

Answer 1

试试这个：

library(tidyverse)
#Data
df <- structure(list(subject = c(1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 3L, 
3L, 3L), object = c("A", "A", "B", "B", "B", "B", "A", "B", "A", 
"B", "B"), value1 = c(TRUE, TRUE, TRUE, TRUE, TRUE, TRUE, TRUE, 
FALSE, TRUE, FALSE, TRUE), value2 = c(TRUE, TRUE, FALSE, FALSE, 
TRUE, FALSE, TRUE, FALSE, TRUE, TRUE, TRUE), value3 = c(FALSE, 
TRUE, TRUE, TRUE, TRUE, FALSE, FALSE, FALSE, FALSE, FALSE, TRUE
)), class = "data.frame", row.names = c(NA, -11L))

#Code
df %>% mutate(value=paste(value1,value2,value3,sep = '.')) %>% group_by(subject,object,value) %>%
  summarize(N=n()) %>% ungroup() %>% group_by(subject,object) %>% mutate(Prop=N/sum(N))

# A tibble: 10 x 5
# Groups:   subject, object [6]
   subject object value                 N  Prop
     <int> <chr>  <chr>             <int> <dbl>
 1       1 A      TRUE.TRUE.FALSE       1 0.5  
 2       1 A      TRUE.TRUE.TRUE        1 0.5  
 3       1 B      TRUE.FALSE.TRUE       2 0.667
 4       1 B      TRUE.TRUE.TRUE        1 0.333
 5       2 A      TRUE.TRUE.FALSE       1 1    
 6       2 B      FALSE.FALSE.FALSE     1 0.5  
 7       2 B      TRUE.FALSE.FALSE      1 0.5  
 8       3 A      TRUE.TRUE.FALSE       1 1    
 9       3 B      FALSE.TRUE.FALSE      1 0.5  
10       3 B      TRUE.TRUE.TRUE        1 0.5

Answer 2

你可以这样做：

 df%>%
   group_by(subject, object)%>%
   mutate(val = str_c(value1,value2,value3,sep = "."),
          value = c(prop.table(table(val))[val]))
# A tibble: 11 x 7
# Groups:   subject, object [6]
   subject object value1 value2 value3 val               value
     <int> <chr>  <lgl>  <lgl>  <lgl>  <chr>             <dbl>
 1       1 A      TRUE   TRUE   FALSE  TRUE.TRUE.FALSE   0.5  
 2       1 A      TRUE   TRUE   TRUE   TRUE.TRUE.TRUE    0.5  
 3       1 B      TRUE   FALSE  TRUE   TRUE.FALSE.TRUE   0.667
 4       1 B      TRUE   FALSE  TRUE   TRUE.FALSE.TRUE   0.667
 5       1 B      TRUE   TRUE   TRUE   TRUE.TRUE.TRUE    0.333
 6       2 B      TRUE   FALSE  FALSE  TRUE.FALSE.FALSE  0.5  
 7       2 A      TRUE   TRUE   FALSE  TRUE.TRUE.FALSE   1    
 8       2 B      FALSE  FALSE  FALSE  FALSE.FALSE.FALSE 0.5  
 9       3 A      TRUE   TRUE   FALSE  TRUE.TRUE.FALSE   1    
10       3 B      FALSE  TRUE   FALSE  FALSE.TRUE.FALSE  0.5  
11       3 B      TRUE   TRUE   TRUE   TRUE.TRUE.TRUE    0.5

计算数据框中值之间交互的比例

calculate proportion of interactions between values in dataframe

r

dataframe

data-wrangling