Filter years with observations of all categories in a variable using dplyr

Question

I have a dataset with rows being individual observations of year and stage, and where there can be zero to multiple observations of a given stage in a given year:

df <- data.frame(year = c(2000, 2000, 2000, 2000, 2001, 2001, 
                          2001, 2002, 2002, 2003, 2003, 2003),
                 stage = c("a", "a", "a", "b", "b", "b",
                           "b", "a", "b", "a", "a", "a")) 
df
##    year stage
## 1  2000     a
## 2  2000     a
## 3  2000     a
## 4  2000     b
## 5  2001     b
## 6  2001     b
## 7  2001     b
## 8  2002     a
## 9  2002     b
## 10 2003     a
## 11 2003     a
## 12 2003     a

I want filter the data to select only the years for which there are observations of both stages a and b (in this case years 2000 and 2002). I have figured out the following way to do this with dplyr and tidyr :

library(dplyr) 
library(tidyr) 

yrs <- df %>% 
  group_by(year, stage) %>%
  summarise(n = n()) %>%
  spread(stage, -year) %>% 
  na.omit %>% 
  pull(year) 

yrs
## [1] 2000 2002

filter(df, year %in% yrs)
##   year stage
## 1 2000     a
## 2 2000     a
## 3 2000     a
## 4 2000     b
## 5 2002     a
## 6 2002     b

This seems a bit clunky and might not scale up well for very large datasets. Is there any simpler, more straightforward way to subset these years using dplyr without calling tidyr::spread ?

Answer 1

You can use group_by %>% filter ; For each group, use all(c('a', 'b') %in% stage) to check if both a and b are within the stage column, and filter the group based on it:

df %>% group_by(year) %>% filter(all(c('a', 'b') %in% stage))

# A tibble: 6 x 2
# Groups:   year [2]
#   year  stage
#  <dbl> <fctr>
#1  2000      a
#2  2000      a
#3  2000      a
#4  2000      b
#5  2002      a
#6  2002      b

Answer 2

Maybe this will work for you:

df %>% group_by(year) %>% 
       filter(length(unique(stage)) == 2)