CODEWITHSUNDEEP
rmagrittr
Jacob
Jacob

Reputation: 422

How to pipe to if statements in R

I have data from my Facebook, Twitter, Instagram, Youtube, and LinkedIn accounts that I'd like to analyze. I have a data frame similar to the following:

df <- data.frame(tw_likes = c(5,4,6,NA,NA,NA,NA,NA,NA), 
                 tw_comments = c(3,5,NA,NA,NA,NA,NA,NA,NA), 
                 fb_likes = c(NA,NA,NA,7,4,8,NA,NA,NA), 
                 fb_comments = c(NA,NA,NA,NA,NA,7,NA,NA,NA), 
                 ig_likes = c(NA,NA,NA,NA,NA,NA,NA,NA,5), 
                 ig_comments = c(NA,NA,NA,NA,NA,NA,43,4,2))

what I want to do is create an additional column Platform that will take the values of "Twitter, "Facebook, or "Instagram" based on the above dataframe.

My tactic has been the following:

for(i in 1:nrow(df){
     if(!is.na(df$tw_likes[i]) | !is.na(df$tw_comments[i])){
          df$Platform[i] <- "Twitter"
     }
     else if(!is.na(df$fb_likes[i]) | !is.na(df$fb_comments[i])){
          df$Platform[i] <- "Facebook"
     }
     else if(!is.na(df$ig_likes[i]) | !is.na(df$ig_comments[i])){
          df$Platform[i] <- "Instagram"
     }
}

This does work, but becomes messier to read. In reality I have more columns and more social media platforms to deal with, so is there a way to pipe the data so I at least don't have to write df$ so many times?

Another thought I had was if I couldn't remove the df$s, could I combine the !is.na() statements to be one statement per if statement?

Upvotes: 2

Views: 997

Answers (3)

Jon Spring
Jon Spring

Reputation: 66425

Here's another approach using dplyr and tidyr to pull the data into long format, filter out the blanks, and add the longer name based on a lookup table:

library(tidyr); library(dplyr)
df %>%
  pivot_longer(cols = everything(), 
               names_to = c("pltfm", "stat"),
               names_sep = "_",
               values_to = "value") %>%
  filter(!is.na(value)) %>%
  left_join(
    tibble(pltfm = c("tw", "fb", "ig"),
           Platform = c("Twitter", "Facebook", "Instagram"))
  )


#Joining, by = "pltfm"
## A tibble: 13 x 4
#   pltfm stat     value Platform 
#   <chr> <chr>    <dbl> <chr>    
# 1 tw    likes        5 Twitter  
# 2 tw    comments     3 Twitter  
# 3 tw    likes        4 Twitter  
# 4 tw    comments     5 Twitter  
# 5 tw    likes        6 Twitter  
# 6 fb    likes        7 Facebook 
# 7 fb    likes        4 Facebook 
# 8 fb    likes        8 Facebook 
# 9 fb    comments     7 Facebook 
#10 ig    comments    43 Instagram
#11 ig    comments     4 Instagram
#12 ig    likes        5 Instagram
#13 ig    comments     2 Instagram

Upvotes: 2

Fino
Fino

Reputation: 1784

Here's an option with dplyr's case_when()

df %>% 
  mutate(Plataform = case_when(
    !is.na(tw_likes) | !is.na(tw_comments) ~ "Twitter",
    !is.na(fb_likes) | !is.na(fb_comments) ~ "Facebook",
    !is.na(ig_likes) | !is.na(ig_comments) ~ "Instagram"))

Upvotes: 4

akrun
akrun

Reputation: 887048

Here is one way in base R to split the dataset into a list of same prefix columns (by removing the suffix substring from the column names), do a rowSums to create a logical matrix, apply max.col to get the column position for each row and change that index by passing a vector of replacement values in the same order of split column names

i1 <- max.col(sapply(split.default(df, sub("_.*", "", names(df))),
        function(x) rowSums(!is.na(x)) > 0 ), 'first')
df$Platform <- c("Facebook", "Instagram", "Twitter")[i1]
df$Platform
#[1] "Twitter"   "Twitter"   "Twitter"   "Facebook"  "Facebook"  
#[6]   "Facebook"  "Instagram" "Instagram" "Instagram"

Upvotes: 4

Related Questions