I had a longer TimeSeries and turned it into wider for forecasting purposes, currently timeseries has the following structure :

I want a function that :

• given a TimeSeries
• loops through columns if cell == NA and previously only NAs were found in the column , keep NA
• if cell != NA good
• if cell == NA But previously we found not NA values, change to previously found value

This would be result :

I tried this function but it isn't right :

filler <- function(df) {
  col <- colnames(df)
  one <- NA
  for (i in col) {
    for (a in i) {
      if(!is.na(a)) {
        one = a
      }
      if(!is.na(one) & is.na(a)) {
        a = one
      }
    }
  }
} 

CodePudding user response：

You may use tidyr::fill -

filler <- function(data) tidyr::fill(data, dplyr::everything())
filler(df)

#    Day Value Strength1 Strength2 Strength3
#1   1/2 1.356         3        NA        NA
#2   2/2 1.385         3        NA        NA
#3   3/2 1.385         3      1.01        NA
#4   4/2 1.400         3      1.01        10
#5   5/2 1.600         3      1.01        10
#6   6/2 1.700         4      1.01        10
#7   7/2 1.800         4      1.05        10
#8   8/2 1.880         4      1.05        10
#9   9/2 1.980         4      1.05        11
#10 10/2 1.800         4      1.05        11