Question

I have a dataframe with three columns: Id, Date and Value and want to downsample this by average: take the next 20 rows, build average of Value from these 20 rows and add it to a new dataframe with the same structure. Date should be the first value of the 20 rows.

I tried it this way (probably horrible :):

resample.downsample <- function(data, by=20)
{
  i <- 0
  nmax <- nrow(data)
  means <- c()
  while(i < nmax)
  {
    means <- c(means, mean(subset(data, Id > i & Id <= i+by)$Value))
    i <- i+by
  }
  return (
    data.frame(
      Id = seq(1, length.out=(nmax/by), by=1),
      Date = seq(startDate, length.out=(nmax/by), by=(1/by)), 
      Value = means  
    )
  )
}

This works so for small datasets, but runs forever on my real datasets (~4000000 rows). Any ideas how to optimize this function?

Sample-Data (input, output should have the same structure, classes: integer, numeric, POSIXct/POSIXt):

    Value   Id  Date
1   125 1   2011-06-30 22:41:50
2   127 2   2011-06-30 22:41:50
3   126 3   2011-06-30 22:41:50
4   123 4   2011-06-30 22:41:50
5   130 5   2011-06-30 22:41:50
6   131 6   2011-06-30 22:41:50
7   128 7   2011-06-30 22:41:50

Answer 1

See this answer for a method that should work for you. How to get the sum of each four rows of a matrix in R. In your case it would be:

colMeans(matrix(data$Value, nrow=20))

Your current method to get the first Date should be fine.