Apply R-function to rows depending on value in other column

Question

I have the following function to build a stock effect for a variable in one column. The variable creates a value in Column B that takes the value in ColumnA and adds a carry over (like e.g. 0.5) from the previous observation in Column B.

constructZ <- function(lag, N) {
  r <- lag^(seq_len(N)-1)
  m <- matrix(rep(r,N),nrow=N)
  z <- matrix(0,nrow=N,ncol=N)
  z[lower.tri(z,diag=TRUE)] <- m[row(m) <= (N+1-col(m))]
  z
}

My problem is now that I have a panel data set that has in one column observations for many different cases. Each case has a specific indicator (numeric). Data looks like:

ColumnA      Indicator         Time
1            1                 1
0            1                 2
0            1                 3
4            2                 1
5            2                 2
0            2                 3
4            3                 1
0            3                 2
2            3                 3

I now want the function to be applied to each case (Indicator) for all observations (Time).

Any idea how to achieve this? The Output should then look like:

ColumnA      Indicator         Time          ColumnB
    1            1                 1         1
    0            1                 2         0.5
    0            1                 3         0.25
    4            2                 1         4
    5            2                 2         7
    0            2                 3         3.5
    4            3                 1         4
    0            3                 2         2
    2            3                 3         3

Any help or support is highly appreciated!

Many thanks in advance!

Answer 1

Here is an alternative loop-free/functional programming solution. We are going to use the Reduce() function which applies a binary function over every pair of items in a vector.

For example, Reduce(`+`, xs) computes the sum of values in vector. If we set accumulate = TRUE, we get a rolling/cumulative sum.

Reduce(`+`, 1:6)
#> [1] 21

# What Reduce is doing here, basically
((((((1) + 2) + 3) + 4) + 5) + 6)
#> [1] 21

# Keep each intermediate sum
Reduce(`+`, 1:6, accumulate = TRUE)
#> [1]  1  3  6 10 15 21

(The purrr package separates these two behaviors into different functions: reduce() and accumulate().)

We can use Reduce() to implement the carry-over/scaling function. First, define a function that works on a pair of values, then use Reduce() to perform a rolling version of it.

rolling_scale <- function(xs, scale_factor) {
  scale_pair <- function(x1, x2) x2 + scale_factor * x1
  Reduce(scale_pair, xs, accumulate = TRUE)
}

rolling_scale(c(4, 5, 0), .5)
#> [1] 4.0 7.0 3.5

Now, we can use dplyr and apply this rolling function to each indicator group.

library(dplyr)

raw <- data.frame(
  ColumnA = c(1, 0, 0, 4, 5, 0, 4, 0, 2), 
  Indicator = rep(x = 1:3, each = 3), 
  Time = 1:3)

raw %>% 
  group_by(Indicator) %>% 
  mutate(ColumnB = rolling_scale(ColumnA, .5)) %>% 
  ungroup()
#> # A tibble: 9 × 4
#>   ColumnA Indicator  Time ColumnB
#>     <dbl>     <int> <int>   <dbl>
#> 1       1         1     1    1.00
#> 2       0         1     2    0.50
#> 3       0         1     3    0.25
#> 4       4         2     1    4.00
#> 5       5         2     2    7.00
#> 6       0         2     3    3.50
#> 7       4         3     1    4.00
#> 8       0         3     2    2.00
#> 9       2         3     3    3.00

Answer 2

Here is my code.

library(dplyr) # Optional, but makes code cleaner
raw = data.frame(ColumnA =
                   c(1, 0, 0, 4, 5, 0, 4, 0, 2),
                 Indicator =
                   rep(x = 1:3, each = 3),
                 Time = 1:3)
factor = 0.5
loop = function(vec) {
  length = length(x = vec)
  if (length == 1) {
    return(vec)
  }
  if (length == 2) {
    return(vec + c(0, vec[2] * factor))
  }
  for (idx in 2:length) {
    vec[idx] = vec[idx] + vec[idx - 1] * factor
  }
  return(vec)
}
output = raw %>%
  mutate(ColumnB = 
           tapply(X = ColumnA,
                  INDEX = Indicator,
                  FUN = loop) %>%
  unlist())
output