[Rd] stats/HoltWinters.R inverted logic in seasonal in R and C

[Norbert Kuder]

Hello,

I have noticed a potentially confusing implementation in the HoltWinters
function regarding the seasonal parameter mapping between R and C code:
https://github.com/wch/r-source/blob/4a1ed749271c52e60a85e794e6f34b0831efb1ae/src/library/stats/R/HoltWinters.R#L98
The C code interprets a seasonal value of 1 as additive and 0 as
multiplicative.
The R seasonal can be "additive" or "multiplicative", so the R code must
invert the logic when calling C.

The proposed solution is to define a seasonalChoice variable:

hw <- function(x, alpha, beta, gamma, seasonal, start.time, f, ...) {
    lenx <- length(x)
    seasonalChoice <- if (seasonal == "multiplicative") 0L else 1L

    .C(C_HoltWinters,
        as.double(x),
        lenx,
        as.double(max(min(alpha, 1), 0)),
        as.double(max(min(beta, 1), 0)),
        as.double(max(min(gamma, 1), 0)),
        as.integer(start.time),
        as.integer(seasonalChoice),
        as.integer(f),
        as.integer(!is.logical(beta) || beta),
        as.integer(!is.logical(gamma) || gamma),
        ...
    )
}

Please check the proposed solution.

Regards,
Norbert Kuder

	[[alternative HTML version deleted]]