CODEWITHSUNDEEP
rconditional-statements
thisisrg
thisisrg

Reputation: 606

Conditional subsetting of a data frame R

Let the data frame be: 

set.seed(123)
df<-data.frame(name=sample(LETTERS,260,replace=TRUE),
               hobby=rep(c("outdoor","indoor"),260),chess=rnorm(1:10))

and the condition which I will use to extract from df be:

df_cond<-df %>% group_by(name,hobby) %>%
    summarize(count=n()) %>%
    mutate(sum.var=sum(count),sum.name=length(name)) %>%
    filter(sum.name==2) %>%
    mutate(min.var=min(count)) %>%
    mutate(use=ifelse(min.var==count,"yes","no")) %>%
    filter(grepl("yes",use))

I want to randomly extract the rows from df that correspond to the (name,hobby,count) combination in df_cond along with the rest of df. I am having bit of a trouble combining %in% and sample.Thanks for any clue!

Edit: For example:

head(df_cond)
         name   hobby count sum.var sum.name min.var   use
      <fctr>  <fctr> <int>   <int>    <int>   <int> <chr>
    1      A  indoor     2       6        2       2   yes
    2      B  indoor     8      16        2       8   yes
    3      B outdoor     8      16        2       8   yes
    4      C outdoor     6      14        2       6   yes
    5      D  indoor    10      24        2      10   yes
    6      E outdoor     8      18        2       8   yes

Using the above data frame, I want to randomly extract 2 rows (=count) with the combination A+indoor(row1) from df, 8 rows with the combination B+indoor (row 2) from df ....and so on.

Combining @denrous and @Jacob answers to get what I need. like so:

m2<-df_cond %>% 
    mutate(data = map2(name, hobby, function(x, y) {df %>% filter(name == x, hobby == y)})) %>% 
    ungroup() %>% 
    select(data) %>% 
    unnest()



test<-m2 %>% 
group_by(name,hobby) %>% 
summarize(num.levels=length(unique(hobby))) %>% 
ungroup() %>% 
group_by(name) %>% 
summarize(total_levels=sum(num.levels)) %>% 
filter(total_levels>1)

fin<-semi_join(m2,test)

Upvotes: 1

Views: 206

Answers (3)

denrou
denrou

Reputation: 640

If I understand correctly, you could use purrr to achieve what you want:

df_cond %>% 
  mutate(data = map2(name, hobby, function(x, y) {filter(df, name == x, hobby == y)})) %>% 
  mutate(data = map2(data, count, function(x, y) sample_n(x, size = y)))

And if you want the same form as df:

df_cond %>% 
  mutate(data = map2(name, hobby, function(x, y) {df %>% filter(name == x, hobby == y)})) %>% 
  mutate(data = map2(data, count, function(x, y) sample_n(x, size = y))) %>% 
  ungroup() %>% 
  select(data) %>% 
  unnest()

Upvotes: 3

Jacob
Jacob

Reputation: 3587

Edited based on OP clarification.

There has to better way but I'd use a loop:

library(dplyr)

master_df <- data.frame()

for (i in 1:nrow(df_cond)){
  name = as.character(df_cond[i, 1])
  hobby = as.character(df_cond[i, 2])
  n = as.numeric(df_cond[i, 3])

  temp_df <- df %>% filter(name == name, hobby == hobby)
  temp_df <- sample_n(temp_df, n)
  master_df <- rbind(master_df, temp_df)
      }

Upvotes: 1

Anand
Anand

Reputation: 3770

Not clear if this is exactly what you want, but you may be looking for left_join:

df %>% 
    left_join(df_cond, by = "name")

Upvotes: 0

Related Questions