R - Faster Way to Calculate Rolling Statistics Over a Variable Interval

Question

I\'m curious if anyone out there can come up with a (faster) way to calculate rolling statistics (rolling mean, median, percentiles, etc.) over a variable interval of time (

Answer 1

In reply to my question to "Kevin" above, I think I figured something out below.

This function takes ticks data (time observations come in at random intervals are and indicated by a time stamp) and calculates the the mean over an interval.

library(Rcpp)

cppFunction('
  NumericVector rollmean_c2( NumericVector x, NumericVector y, double width,
                              double Min, double Max) {

double total = 0, redge,center;
unsigned int n = (Max - Min) + 1,
                  i, j=0, k, ledge=0, redgeIndex;
NumericVector out(n);


for (i = 0; i < n; i++){
  center = Min + i + 0.5;
  redge = center - width / 2;
  redgeIndex = 0;
  total = 0;

  while (x[redgeIndex] < redge){
    redgeIndex++;
  }
  j = redgeIndex;

  while (x[j] < redge + width){
    total += y[j++];

  }

  out[i] = total / (j - redgeIndex);
}
return out;

  }')

# Set up example data
x = seq(0,4*pi,length.out=2500)
y = sin(x) + rnorm(length(x),0.5,0.5)
plot(x,y,pch=20,col="black",
     main="Sliding window mean; width=1",
     sub="rollmean_c in red      rollmean_r overlaid in white.")


c.out = rollmean_c2(x,y,width=1,Min = min(x), Max = max(x)) 
lines(0.5:12.5,c.out,col="red",lwd=3)