Mutate something on complete cases, but keep all-CodePudding

I want to generate a group id, based on combinations of two variables (country and party). This is my data:

df <- data.frame(country = c("BE", "BE", "BE", "NL", "NL", "NL"),
                 year = c(2010, 2010, 2010, 2010, 2010, 2010),
                 party = c(NA, NA, NA, "A", "B", "B"))

Which gives:

  country year party
1      BE 2010  <NA>
2      BE 2010  <NA>
3      BE 2010  <NA>
4      NL 2010     A
5      NL 2010     B
6      NL 2010     B

What I want is:

  country  year party group
  <chr>   <dbl> <chr> <int>
1 BE       2010 NA        NA
2 BE       2010 NA        NA
3 BE       2010 NA        NA
4 NL       2010 A         1
5 NL       2010 B         2
6 NL       2010 B         2

I tried:

df <- df %>% 
  group_by(country, party) %>% 
  mutate(group = cur_group_id())

But this gives me:

  country  year party group
  <chr>   <dbl> <chr> <int>
1 BE       2010 NA        1
2 BE       2010 NA        1
3 BE       2010 NA        1
4 NL       2010 A         2
5 NL       2010 B         3
6 NL       2010 B         3

However, I don't want separate groups for any data that has missing values. At the same time, I want to keep the data.

If I try:

df <- df %>% 
  group_by(country, party) %>% 
  filter(!is.na(party)) %>% 
  mutate(group = cur_group_id())

I get:

  country  year party group
  <chr>   <dbl> <chr> <int>
1 NL       2010 A         1
2 NL       2010 B         2
3 NL       2010 B         2

How can I get this new variable only for complete data, whilst keeping the incomplete data in the dataset?

Thanks

CodePudding user response：

Use interaction

df %>% mutate(group = as.integer(interaction(country, party, drop = TRUE)))

giving:

  country year party group
1      BE 2010  <NA>    NA
2      BE 2010  <NA>    NA
3      BE 2010  <NA>    NA
4      NL 2010     A     1
5      NL 2010     B     2
6      NL 2010     B     2

CodePudding user response：

df <- data.frame(country = c("BE", "BE", "BE", "NL", "NL", "NL"),
                 year = c(2010, 2010, 2010, 2010, 2010, 2010),
                 party = c(NA, NA, NA, "A", "B", "B")) 

library(data.table)
setDT(df)[!is.na(party), grp := .GRP, by = party][]
#>    country year party grp
#> 1:      BE 2010  <NA>  NA
#> 2:      BE 2010  <NA>  NA
#> 3:      BE 2010  <NA>  NA
#> 4:      NL 2010     A   1
#> 5:      NL 2010     B   2
#> 6:      NL 2010     B   2

^{Created on 2021-12-21 by the reprex package (v2.0.1)}

CodePudding user response：

Something like the following?

library(tidyverse)

df <- data.frame(country = c("BE", "BE", "BE", "NL", "NL", "NL"),
                 year = c(2010, 2010, 2010, 2010, 2010, 2010),
                 party = c(NA, NA, NA, "A", "B", "B")) 

df %>% 
  group_by(country, party) %>% 
  mutate(group = if_else(is.na(party), NA_integer_, cur_group_id()))
#> # A tibble: 6 × 4
#> # Groups:   country, party [3]
#>   country  year party group
#>   <chr>   <dbl> <chr> <int>
#> 1 BE       2010 <NA>     NA
#> 2 BE       2010 <NA>     NA
#> 3 BE       2010 <NA>     NA
#> 4 NL       2010 A         2
#> 5 NL       2010 B         3
#> 6 NL       2010 B         3

If you want the groups to begin with 1 (instead of 2):

library(tidyverse)

df %>% 
  filter(!is.na(party)) %>% 
  group_by(country, party) %>% 
  mutate(group = cur_group_id()) %>% 
  ungroup %>% add_row(filter(df,is.na(party))) %>% 
  mutate(group = if_else(is.na(party), NA_integer_, group))

#> # A tibble: 6 × 4
#>   country  year party group
#>   <chr>   <dbl> <chr> <int>
#> 1 NL       2010 A         1
#> 2 NL       2010 B         2
#> 3 NL       2010 B         2
#> 4 BE       2010 <NA>     NA
#> 5 BE       2010 <NA>     NA
#> 6 BE       2010 <NA>     NA