CODEWITHSUNDEEP
rfilterdplyr
conor
conor

Reputation: 1324

R dplyr filter requires stored filter condition

When I filter a dataframe where the condition is made up of other filters, it doesn't seem to work. However, if I store the condition as a variable (f in the example), the filtering works fine. Can someone explain why this happens, and how to make something like Example 2 work? I would prefer to not store the filter condition as a variable.

library(dplyr)

# Dummy data set
df <- data.frame(Country = factor(c("Argentina", "Brazil", "Brazil", "Brazil")), 
                 Type = factor(c("A", "A", "B", "C")))

# Only returns Brazil. No problem here.
f <- df %>% 
  group_by(Country) %>% 
  summarise(nTypes = n_distinct(Type)) %>% 
  filter(nTypes==3) %>% 
  select(Country) %>% 
  droplevels() %>% 
  unlist()
# > f
#   Country 
# Brazil 
# Levels: Brazil


# Example 1 - Only returns rows of df where Country=="Brazil". No problem here.
df %>% filter(
  Country %in% (f
                )
  )
#   Country Type
# 1  Brazil    A
# 2  Brazil    B
# 3  Brazil    C


# Example 2 - Filter is equivalent to `f` but returns all rows of df, not just Brazil. No idea why!
df %>% filter(
  Country %in% (df %>% 
                  group_by(Country) %>% 
                  summarise(nTypes = n_distinct(Type)) %>% 
                  filter(nTypes==3) %>% 
                  select(Country) %>% 
                  droplevels() %>% 
                  unlist()
                )
  )
#     Country Type
# 1 Argentina    A
# 2    Brazil    A
# 3    Brazil    B
# 4    Brazil    C

Upvotes: 1

Views: 295

Answers (1)

amrrs
amrrs

Reputation: 6325

While I'm not sure why you are getting unexpected results, based on this answer: Using filter inside filter in dplyr gives unexpected results a way to get desired result after filtering is to use inner_join

df %>% 
  group_by(Country) %>% 
  summarise(nTypes = n_distinct(Type)) %>% 
  filter(nTypes==3) %>% 
  select(Country) %>% inner_join(.,df)

Output:

Joining, by = "Country"
# A tibble: 3 x 2
  Country   Type
   <fctr> <fctr>
1  Brazil      A
2  Brazil      B
3  Brazil      C

Upvotes: 1

Related Questions