R: Group average of all nearby points in x and y coordinates -

i have data involves success/fail variable many different locations x-y coordinates (integers between 0 , 80 both). want model expected probability @ particular location. can pretty using plyr:

sucprop <- ddply(df, .(xcrd,ycrd), function(x) data.frame(obs=nrow(x),prop=mean(x$success))) 

this gives me proportion of successes @ each coordinate. success rate @ 1 point should similar @ nearby points wondering how can best take average successes of points within +-5 in both x , y direction.

so (25,50) point take average of observations within (20-30,45-55).

what best way this? can input straight .variables in ddply or have work out sort of rolling index?

you have rolling mean of sorts. here's 1 way sapply:

data.frame(     x = rnorm(10, 40, 5),     y = rnorm(10, 50, 7),     success = rbinom(10, 1, .4)) -> ff  newmean <- function(q) {     ff[q,"x"] + 5 ->     ff[q,"x"] - 5 -> b     ff[q,"y"] + 5 -> c     ff[q,"y"] - 5 -> d     ff[ff$x < & ff$x >b & ff$y < c & ff$y > d, "success"] -> k     mean(k) }  sapply(x = 1:nrow(ff), newmean) -> ff$neighborhood_prob