proxy package in R, can't make it work

Question

I am wanting to implement an error-weighted euclidean distance function in R (similar but not quite the same as Kumar & Patel. 2005. Rutcor Research Report. RRR 12-2005). I already know that I want to use pr_DB within proxy for this. Here is where I run into problems. The first thing I always do when looking at a new package is cut-and-paste the examples into R to see if they do what I expect them to do. This is not what happened in this case. I started with the simple included example, to wit:

mydist <- function(x,y) x * y
pr_DB$set_entry(FUN = mydist, names = c("test", "mydist"))

Okay, so far, so good. I figure, if it's going to work just like a normal "dist" type function, I'll toss in a very simple piece of data.

>toydat
        x           y
a -0.12817993 -1.03238513
b  1.56200731  0.93826937
c -1.24051847 -1.31005852
d -1.12892553 -1.57133401
e -1.10098308  0.06577006

Just a little toy set, of course. SOOOOOO, I try out the following:

toydist <- dist(toydat,method="mydist")

I get the following message:

Error in do.call(".External", c(list(CFUN, x, y, pairwise, if (!is.function(method)) get(method) else method),  : 
not a scalar return value

On a hunch, I tried out:

toydist <- dist(toydat,method="Euclidean")
toydist <- dist(toydat,method="Manhatten")

and others. They all work as expected. I am presuming that there is something special that must be done with the basic formula for it to work in a proper manner for computing distance matrices. What I want to compute is sqrt((x_i - x_j)²) + (y_i - y_j)²) + ...(n_i - n_j)²) ÷ sqrt((σx_i² + σx_j²) + (σy_i² + σy_j²) + ...(σn_i² + σn_j²)), in every pairwise combination for my data set.

I realize I could do this by spreadsheet for a single data set, but I want to do some clustering bootstrapping, which I know how to do in R. Not practical with a spreadsheet.

Answer 1

Now that the requisite time has passed, I can close this out formally. The function I came up with is thus:

DistErrAdj <- function(x,y) {
sing.err <- sqrt((x^2) + (y^2))
sum(sing.err)
}