CODEWITHSUNDEEP
rfor-loopdplyr
Arthur Carvalho Brito
Arthur Carvalho Brito

Reputation: 560

Place a for loop in the middle of a pipe in R

I would like some help to make my code a bit leaner.

Here is a fraction of my original dataset:

library(tidyverse)
dataset <- data.frame(UPA = c(130033353, 130033353, 130033353, 130033353,
                              130033353, 230036930, 230036930, 230036930, 230036930, 230036930,
                              230124582, 230124582, 230124582, 230124582, 230124582, 240039107,
                              240039107, 240039107, 240039107, 240039107, 320022393, 320022393,
                              320022393, 320022393, 320022393, 330093898, 330093898, 330093898,
                              330093898, 330093898),
                      UF = c(13, 13, 13, 13, 13, 23, 23, 23,23, 23, 23, 23, 23, 23, 23, 24, 24,
                             24, 24, 24, 32, 32, 32, 32, 32, 33, 33, 33, 33, 33),
                      V1008 = c(1, 1, 1, 1, 1, 5, 5, 5, 5,5, 11, 11, 11, 11, 11,
                                8, 8, 8, 8, 8, 3, 3, 3, 3, 3, 9, 9, 9,9, 9),
                      V1014 = c(2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
                                2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2),
                      V2008 = c(22, 23, 12, 28, 29, 13, 9, 9, 2, 22, 18, 5, 8,
                                16, 16, 12, 24, 2, 25, 28, 7, 7, 7, 15, 15, 6,
                                6, 18, 14, 14),
                      V20081 = c(1, 9, 6, 3, 5, 9, 7, 6, 9, 5, 2, 6, 1, 5, 9,
                                 10, 5, 4, 5, 1, 7, 7, 7, 5, 5, 8, 8, 4, 8, 8),
                      V20082 = c(1952, 1964, 1995, 1999, 2009, 1993, 1998, 2000,
                                 2003, 2011, 1967, 1990, 1993, 1996, 2001, 1947, 1996,
                                 1998, 1997, 2012, 2010, 2010, 2010, 2011, 2011, 1981, 1981, 1984,
                                 2006, 2006),
                      V2003 = c(1, 2, 6, 8, 10, 2, 9, 10, 12, 15, 2, 3,
                                5, 7, 11, 2, 6, 7, 7, 9, 3, 3, 3, 4, 4, 1, 1, 2, 4, 4),
                      V2007 = c(1, 2, 2, 1, 2, 2, 1, 1, 1, 1, 2, 1, 2, 1, 1, 2, 2, 1, 2,
                                2, 1, 1, 1, 1, 1, 2, 2, 1, 2, 2),
                      n_p = c(1, NA, NA, NA, NA, NA, NA, NA,
                              NA, NA, NA, NA, NA, NA, NA, NA, 
                              NA, NA, NA, NA, NA, NA, NA, NA,
                              NA, 1, 2, 1, NA, NA),
                      id_dom = c(2499L, 2499L, 2499L, 2499L,
                                 2499L, 10962L, 10962L, 10962L, 10962L, 
                                 10962L, 12618L, 12618L, 12618L, 12618L, 12618L, 13673L, 
                                 13673L, 13673L, 13673L, 13673L, 25945L, 25945L, 25945L, 
                                 25945L, 25945L, 28145L, 28145L, 28145L, 28145L, 28145L))

I want to update the value of the p201 variable according to n_p's value in an iterative way. n_p can go possibly from 1 to 5 (although not in dataset, I apologize for not being able to take a sample that contains all possible cases). In the first wave, I assign p201 to be equal to 100*(n_p-1) + V2003 if n_p == 1 and NA otherwise. After that, I go from 2 to 5 doing the same considering only rows in which p201 is still missing.

Here is the code: 

final_df <- dataset %>%
  group_by(UF, UPA, V1008, V1014, V2007, 
           V2008, V20081, V20082, V2003) %>%
  mutate(p201 = ifelse(n_p == 1 & V2008 != 99 &
                         V20081 != 99 & V20082 != 9999,
                       100*(n_p-1) + V2003, NA)) %>%
    fill(p201, .direction = 'down') %>%
    mutate(p201 = ifelse(n_p == 1,
                         p201,
                         ifelse(n_p == 2 & is.na(p201) & V2008 != 99 &
                                  V20081 != 99 & V20082 != 9999, 
                                100*(n_p-1) + V2003,
                                NA))) %>%
    fill(p201, .direction = 'down') %>%
    mutate(p201 = ifelse(n_p %in% 1:2,
                         p201,
                         ifelse(n_p == 3 & is.na(p201) & V2008 != 99 &
                                  V20081 != 99 & V20082 != 9999,
                                100*(n_p-1) + V2003,
                                NA))) %>%
    fill(p201, .direction = 'down') %>%
    mutate(p201 = ifelse(n_p %in% 1:3,
                         p201,
                         ifelse(n_p == 4 & is.na(p201) & V2008 != 99 &
                                  V20081 != 99 & V20082 != 9999,
                                100*(n_p-1) + V2003,
                                NA))) %>%
    fill(p201, .direction = 'down') %>%
    mutate(p201 = ifelse(n_p %in% 1:4,
                         p201,
                         ifelse(n_p == 5 & is.na(p201) & V2008 != 99 &
                                  V20081 != 99 & V20082 != 9999,
                                100*(n_p-1) + V2003,
                                NA))) %>%
    ungroup() %>%
    mutate_at(c('UF', 'UPA', 'V1008', 'p201'), as.character) %>%
    mutate(idind = ifelse(is.na(p201),
                          NA,
                          paste0(V1014, UF, UPA, V1008, p201)))

I clearly could use something like a for loop (or maybe something even better, using map?) to make the code a bit leaner, but I don't know how to insert a loop inside a pipe flow using tidyverse's syntax.

Can somebody help? My desired output is exactly the one resulting in final_df, but with cleaner code.

PS: Please do not mind that the result will generate a lot of NAs in p201 - the whole dataset is more complex and this probably wouldn't be true with it.

EDIT I've figured out a solution by adapting Limey's answer - I didn't know about recursive functions.

Here it goes: 

loop <- function(data,
                 interview = 2,
                 int_final = 5){
  data <- data %>%
    group_by(UF, UPA, V1008, V1014, V2007,
             V2008, V20081, V20082, V2003) %>%
    fill(p201, .direction = 'down') %>%
    mutate(p201 = ifelse(
      n_p %in% 1:(interview-1),
      p201,
      ifelse(
          n_p == interview  & is.na(p201) &
            V2008 != 99 &
            V20081 != 99 & V20082 != 9999,
          100 * (n_p - 1) + V2003,
          NA
        )
      ))

    if(interview == int_final){
      return(data)
      } else{
        return(loop(data, interview + 1, int_final))
        }
}

final_dataset <- dataset %>%
  group_by(UF, UPA, V1008, V1014, V2007,
           V2008, V20081, V20082, V2003) %>%
  mutate(p201 = ifelse(n_p == 1  & V2008 != 99 &
                         V20081 != 99 & V20082 != 9999,
                       100*(n_p-1) + V2003, NA)) %>%
  loop() %>%
  ungroup() %>%
  mutate_at(c('UF', 'UPA', 'V1008', 'p201'), as.character) %>%
  mutate(idind = ifelse(is.na(p201),
                        NA,
                        paste0(V1014, UF, UPA, V1008, p201)))

Upvotes: 1

Views: 166

Answers (1)

Limey
Limey

Reputation: 12461

Hmmm. Sounds like a but of recursive programming might do the trick. I have no idea why you're trying to do this and your haven't provided a desired outcome, so I can't check my results, but something like this might work.

[Untested code]

doIt <- function(data, currentDepth=1, maxDepth=5) {
  data <- data %>%
    group_by(UF, UPA, V1008, V1014, V2007, 
             V2008, V20081, V20082, V2003) %>%
    mutate(p201 = ifelse(n_p %in% 1:currentDepth & V2008 != 99 & V20081 != 99 & V20082 != 9999,
                         100*(n_p-1) + V2003, 
                         NA)) %>%
    fill(p201, .direction = 'down')
    if (currentDepth == maxDepth) return(data)
    else return (doIt(data, currentDepth+1, maxDepth))
}

final_dataset <- doIt(dataset)

I've tried to generalise your ifelse. The function either calls itself again (if current depth < maxDepth) or returns its own result.

Upvotes: 1

Related Questions