Alternative to na.locf (zoo) for numbers

Question

I currently have a fairly complicated dataset, but I tried to break it down and hope that the problem at hand catches sufficient complexity.

My data looks as follows:

df <- data.frame(c(1,1,1,1,2,2,2,3,3,3), c(3,3,NA,5,0,0,0,7,4,7), 
                 c(TRUE, TRUE, TRUE, FALSE, FALSE, FALSE, FALSE, FALSE,
                   TRUE, FALSE))
colnames(df) <- c("ID", "Number", "Status")

> df
   ID Number Status
1   1      3   TRUE
2   1      3   TRUE
3   1     NA   TRUE
4   1      5  FALSE
5   2      0  FALSE
6   2      0  FALSE
7   2      0  FALSE
8   3      7  FALSE
9   3      4   TRUE
10  3      7  FALSE

I am looking at each ID in turn (using dplyr group_by). Whenever the status of an observation is TRUE, I would like to replace "Number" by the subsequent number with the status FALSE. For ID = 1, this would imply the number 5 for all 4 observations.

I have found a workaround, but I am sure there must be an easier solution (using replace?). This is how I proceeded:

library(dplyr)
library(zoo)

# Setting up a new variable that replaces all "unwanted
# numbers by NA
df$newNumber <- NA
df$newNumber[df$Status == FALSE] <- df$Number[df$Status == FALSE]

# Using the zoo function na.locf to replace the Gas
df <- df %>%
  group_by(ID) %>%
  mutate(Number2 = ifelse(any(Status == TRUE), na.locf(newNumber, 
fromLast = TRUE), Number2))

> df
# A tibble: 10 x 5
# Groups:   ID [3]
      ID Number Status newNumber Number2
   <dbl>  <dbl> <lgl>      <dbl>   <dbl>
 1     1      3 TRUE          NA       5
 2     1      3 TRUE          NA       5
 3     1     NA TRUE          NA       5
 4     1      5 FALSE          5       5
 5     2      0 FALSE          0       0
 6     2      0 FALSE          0       0
 7     2      0 FALSE          0       0
 8     3      7 FALSE          7       7
 9     3      4 TRUE          NA       7
10     3      7 FALSE          7       7

Thank you very much in advance!

dipetkov dipetkov · Accepted Answer · 2019-03-18T17:52:51

tidyr::fill works much the same way as zoo::na.locf.

library("tidyverse")

df <- data.frame(c(1,1,1,1,2,2,2,3,3,3), c(3,3,NA,5,0,0,0,7,4,7),
                 c(TRUE, TRUE, TRUE, FALSE, FALSE, FALSE, FALSE, FALSE,
                   TRUE, FALSE))
colnames(df) <- c("ID", "Number", "Status")

df %>%
  mutate(Number2 = if_else(Status, NA_real_, Number)) %>%
  group_by(ID) %>%
  fill(Number2, .direction = "up")
#> # A tibble: 10 x 4
#> # Groups:   ID [3]
#>       ID Number Status Number2
#>    <dbl>  <dbl> <lgl>    <dbl>
#>  1     1      3 TRUE         5
#>  2     1      3 TRUE         5
#>  3     1     NA TRUE         5
#>  4     1      5 FALSE        5
#>  5     2      0 FALSE        0
#>  6     2      0 FALSE        0
#>  7     2      0 FALSE        0
#>  8     3      7 FALSE        7
#>  9     3      4 TRUE         7
#> 10     3      7 FALSE        7

^{Created on 2019-03-18 by the reprex package (v0.2.1)}

Alternative to na.locf (zoo) for numbers

2 Answers