更新

Question

我正在尝试将两个数据帧合并在一起，这两个数据帧通过名为patient的特定变量相互关联。第二个数据帧具有同一患者列的多个条目。我不想在合并时创建重复的患者条目，但是我想通过将一列下的值连接起来在第二个数据框中保留唯一信息。

我尝试使用有效的group_by手动连接某些变量。我有几个变量，但是手动指定所有变量都不可行

我还可以使用dplyr连接数据帧中的每个变量，如下所示。第二种情况的问题是重复值也被串联在一起，使数据帧不必要地变大且难以处理。请参阅下面的reprex。

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union

df1 <- data.frame(patient=c("a", "b", "c"), 
                  var1 = 1:3, 
                  var2=11:13)

df1
#>   patient var1 var2
#> 1       a    1   11
#> 2       b    2   12
#> 3       c    3   13



df2 <- data.frame(patient=c("a","a", "b", "b", "c", "c" ), 
                  treatment=  rep(c("drug1", "drug2"), 3), 
                  time= rep(c("time1", "time2"), 3),
                  var3= "constant")

df2
#>   patient treatment  time     var3
#> 1       a     drug1 time1 constant
#> 2       a     drug2 time2 constant
#> 3       b     drug1 time1 constant
#> 4       b     drug2 time2 constant
#> 5       c     drug1 time1 constant
#> 6       c     drug2 time2 constant


df_merged <- left_join(df1, df2)
#> Joining, by = "patient"

# Don't want duplicates like this
df_merged
#>   patient var1 var2 treatment  time     var3
#> 1       a    1   11     drug1 time1 constant
#> 2       a    1   11     drug2 time2 constant
#> 3       b    2   12     drug1 time1 constant
#> 4       b    2   12     drug2 time2 constant
#> 5       c    3   13     drug1 time1 constant
#> 6       c    3   13     drug2 time2 constant



df_merged2 <- df_merged %>%
  group_by(patient) %>%
  mutate(treatment = paste(treatment, collapse = "_"),
         time=paste(time, collapse = "_")) %>%
  filter(!duplicated(patient))

# I can manually edit a few variables like this
df_merged2
#> # A tibble: 3 x 6
#> # Groups:   patient [3]
#>   patient  var1  var2 treatment   time        var3    
#>   <fct>   <int> <int> <chr>       <chr>       <fct>   
#> 1 a           1    11 drug1_drug2 time1_time2 constant
#> 2 b           2    12 drug1_drug2 time1_time2 constant
#> 3 c           3    13 drug1_drug2 time1_time2 constant


df_merged3 <- df_merged %>%
  group_by(patient) %>%
  mutate_at(vars(-group_cols()), .funs = ~paste(., collapse ="_")) %>%
  filter(!duplicated(patient))


# I have many variables I can't specify manually
# I can create this merged data frame, but I don't want to 
# concatenate duplicated values such as var1, var2, and var3
df_merged3
#> # A tibble: 3 x 6
#> # Groups:   patient [3]
#>   patient var1  var2  treatment   time        var3             
#>   <fct>   <chr> <chr> <chr>       <chr>       <chr>            
#> 1 a       1_1   11_11 drug1_drug2 time1_time2 constant_constant
#> 2 b       2_2   12_12 drug1_drug2 time1_time2 constant_constant
#> 3 c       3_3   13_13 drug1_drug2 time1_time2 constant_constant

^{由reprex package（v0.3.0）于2019-10-23创建}

我想看看是否有一种方法可以连接仅包含唯一值的变量，以保留第二个数据帧中的信息，而无需复制df_merged中的行。

如果您有除dplyr以外的其他建议，我将很高兴听到。 data.table解决方案也可能也适合我，因为我的实际数据帧很大。

谢谢！

Answer 1

我们可以使用summarise_at和unique

library(dplyr)
df_merged %>%
          group_by(patient) %>%
          summarise_at(vars(-group_cols()), .funs = ~paste(unique(.), collapse ="_"))

或者我们可以直接进行合并/联接，而不用中间数据帧添加/更改Global Env。

left_join(df1, 
          df2 %>% group_by(patient) %>% 
                  summarise_at(vars(-group_cols()), .funs = ~paste(unique(.), collapse ="_")) %>% 
                  ungroup()
         )

Joining, by = "patient"
  patient var1 var2   treatment        time     var3
1       a    1   11 drug1_drug2 time1_time2 constant
2       b    2   12 drug1_drug2 time1_time2 constant
3       c    3   13 drug1_drug2 time1_time2 constant

更新

#Here a toy example to experiment with, uncomment browser to see how it works inside Reduce, 
#also see ?Reduce for more info 
paste_mod <- function(x) Reduce(function(u, v){
u <- ifelse(!grepl('_',u) & is.na(u),'.',u)
v <- ifelse(is.na(v),'.',v)
if(v=='.' | !grepl(v,u)) paste0(u,'_',v) else u
}, x)

paste_mod(c("drug1",NA,NA,"drug2","drug1","drug2"))
[1] "drug1_._._drug2"
paste_mod(c(NA,NA,"drug2","drug1","drug2"))
[1] "._._drug2_drug1"

#replace NA with . then apply Reduce
df2 %>% 
       mutate_if(is.factor,as.character) %>% mutate_all(~replace(.,is.na(.),'.')) %>% 
       group_by(patient) %>% 
       summarise_at(vars(-group_cols()), .funs = ~Reduce(function(u, v) if(v=='.' | !grepl(v,u)) paste0(u,'_',v) else u, .)) %>% 
       ungroup()

# A tibble: 2 x 4
  patient treatment       time        var3    
  <chr>   <chr>           <chr>       <chr>   
1 a       drug1_._._drug2 time1_time2 constant
2 c       drug1_drug2     time1_time2 constant

新的df2用于测试更新的解决方案

df2 <- structure(list(patient = structure(c(1L, 1L, 1L, 1L, 2L, 2L), .Label = c("a", 
"c"), class = "factor"), treatment = structure(c(1L, NA, NA, 
2L, 1L, 2L), .Label = c("drug1", "drug2"), class = "factor"), 
    time = structure(c(1L, 2L, 1L, 2L, 1L, 2L), .Label = c("time1", 
    "time2"), class = "factor"), var3 = structure(c(1L, 1L, 1L, 
    1L, 1L, 1L), class = "factor", .Label = "constant")), class = "data.frame", row.names = c(NA, 
-6L))

连接数据框而不会创建重复的行，同时将一列下的唯一条目串联在一起

1 个答案:

更新