R replacing NAs with increment/decrement of non-NA values

user_12672193 0
Viewed 222

I'm trying to fill in missing year values by a group both incrementally and decrementally when I know that each increment/decrement should be 1. Here's a reproducible example. Note there are two ids where the first id starts in 2002 (although it's missing) and the second id starts in 2007 (again also missing). I know this from the existing year and the fact that the increment/decrement should be 1.

id <- rep(c("01012895", "01021890"), each = 6)
value <- c(rnorm(12))
year <- c("NA", "NA", 2004, "NA", "NA", 2007, "NA", "NA", "NA", 2010, "NA", "NA")
df <- data.frame(id, value, year)

df
         id       value year
1  01012895 -1.05853432   NA
2  01012895  0.06480395   NA
3  01012895 -3.32811206 2004
4  01012895 -0.33427578   NA
5  01012895 -0.22790538   NA
6  01012895  1.46991465 2007
7  01021890 -0.65928777   NA
8  01021890 -0.25701765   NA
9  01021890  1.24331689   NA
10 01021890  0.10006530 2010
11 01021890 -0.83971071   NA
12 01021890  1.15893263   NA

df2 is what I want eventually. 

year2 <- c(2002:2007, 2007:2012)
df2 <- data.frame(id, value, year2)

I've tried to find a get around using fill, like in df3, but it only carries existing values without an option to alter it. 

df3 <- df %>% group_by(site_no) %>% fill(year, .direction = "updown")
dplyr na r replace
1 Answers

Here's one way to do it.

First, your own data:

id <- rep(c("01012895", "01021890"), each = 6)
value <- c(rnorm(12))
year <- c(NA, "NA", 2004, "NA", "NA", 2007, "NA", "NA", "NA", 2010, "NA", "NA")
df <- data.frame(id, value, year)

year2 <- c(2002:2007, 2007:2012)
df2 <- data.frame(id, value, year2)

Converting to integer from factor and using row numbers and grouped max indexes, we get the following:

library(dplyr)

df3 <- df %>% 
  group_by(id) %>%
  mutate(
    year = as.integer(as.character(year)), # Convert factor to integer
    year =  max(year, na.rm = TRUE) + (row_number() - which.max(year))
  )

df3
#> # A tibble: 12 x 3
#> # Groups:   id [2]
#>    id         value  year
#>    <fct>      <dbl> <int>
#>  1 01012895  0.0482  2002
#>  2 01012895 -0.445   2003
#>  3 01012895  0.459   2004
#>  4 01012895 -0.740   2005
#>  5 01012895  1.91    2006
#>  6 01012895 -1.66    2007
#>  7 01021890 -1.08    2007
#>  8 01021890 -0.816   2008
#>  9 01021890  0.395   2009
#> 10 01021890 -0.0294  2010
#> 11 01021890  0.397   2011
#> 12 01021890  0.273   2012

all(df2 == df3)
#> [1] TRUE

It does throw a warning, which should be fine to ignore.

user_118907890
Related