Count observations using rollapply from zoo package with dynamic criteria

Question

I am looking for a way to count observations like here but with the ability to change criteria according to the specific observation (moving count).

For example - count the number of observations of mag (from the last 50) which are greater than the specific observation of mag. The code I have:

rollapplyr(zoo(mag),50,function(i){sum(mag>i)},partial=T,by.column=F,fill=NA))

This code takes the average mag of the 50 last observations and calculates the number of observations above that average (in the whole data-set).

What am I missing ? Maybe using rollapply is not the case here ? To sum it up:
1. count according to specific row value
2. count only in the 50 last observations (and not the whole data-column).

Answer 1

Please see the "corrected" function below:

set.seed(2017)
mag <- sample(x = 1000, size = 20)

## Your function, see what is printed
# my_fun <- function(i) {
#   print(i)
#   print(mag)
#   sum(mag > i)
# }

## Corrected one
my_fun <- function(i) {
  print(i)
  print(tail(i, 1))
  sum(i > tail(i, 1))
}

# debug(my_fun)  # Play a little with debug(), it is worth it!

mag_out <- zoo::rollapplyr(
  # zoo::zoo(mag),
  mag,
  5,
  my_fun,
  partial = TRUE,
  by.column = FALSE,
  fill = NA
)

rbind(
  mag,
  mag_out
)

Output:

        [,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9] [,10] [,11] [,12] [,13] [,14] [,15] [,16] [,17] [,18] [,19] [,20]
mag      244  329  987  833  524  112  869  327  488   691    89   224   206    73   803   868   288   365   666   145
mag_out    0    0    0    1    2    4    1    3    2     1     4     3     3     4     0     0     2     2     2     4