Keith Jewell said:
> ... from reading ?all.equal I would have expected
> scale = 1 and the default scale = NULL to give identical results for the
> length
> one numerics being passed to all.equal.
>
> Can anyone explain?
Inspectng the code in all.equal.numeric, I find
xy <- mean((if (cplx)
Mod
else abs)(target - current))
if (is.null(scale)) {
xn <- mean(abs(target))
if (is.finite(xn) && xn > tolerance) {
xy <- xy/xn
"relative"
}
else "absolute"
}
else {
xy <- xy/scale
if (scale == 1)
"absolute"
else "scaled"
}
target is the first number supplied, current the second; in yoour example code
that is x[2] and x[1] respectively. Later on xy is compared to the tolerance.
In the code, scale=NULL and scale=1 are clearly treated differently; in
particular when scale is NULL the absolute difference is divided by first of
the (two) numbers if that number is greater than tolerance or is used
unchanged, and if scale=1 it is divided by scale throughout.
That would mean that for scale=NULL, your example will divide the difference by
10, 9, ..1 in that order before comparing with tolerance, and if scale=1 it
will simply compare the difference directly with the tolerance. Calculating
your case through for scale = NULL, xy will take the values
ifelse(b>5, abs(a-b)/b, abs(a-b))
[1] 0.9000000 0.7777778 0.6250000 0.4285714 0.1666667 1.0000000 3.0000000
[8] 5.0000000 7.0000000 9.0000000
Of those, only the last 2 are greater than 5, which is the result you found. By
contrast, when scale=1 xy takes the values
abs(a-b)
[1] 9 7 5 3 1 1 3 5 7 9
of which the two at each end are both greater than 5.
That fairly complicated behavio0ur is probably a good reason to use a simpler
calculation in which you can see how the difference is being scaled ... ;)
Steve Ellison
*******************************************************************
This email and any attachments are confidential. Any use...{{dropped:8}}
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
