[Rd] bugs in documentation of stats::stl

Mark van der Loo mark.vanderloo at gmail.com
Wed Aug 23 11:10:50 CEST 2017

Dear list, R-core,

The documentation of stats::stl explicitly refers to the paper by
Cleveland[1] to explain the parameters. However, the description is
confusing, with two descriptions seeming to refer to the same parameter in
the paper.

s.window: [...] the loess window for seasonal extraction, which should be
odd and at least 7, according to Cleveland et al

--> The phrase 'odd and at least 7' refers to Cleveland's parameter n_(s),
section 3.5 of [1].

Confusing: Cleveland calls this 'seasonal smoothing', not extraction.

l.window:  the span (in lags) of the loess window of the low-pass filter
used for each subseries.[...]

--> The description 'low-pass filter used for each subseries' also seems to
correspond to Cleveland's parameter n_(s), in step two of the algorithm in
the reference. (section 2.2 of [1]).

Confusing: Cleveland does not apply a low-pass filter to each subseries[2].
The subseries are reconstructed to a single series and after that a
low-pass filter is applied (step 3 of the algorithm in section 2.2 of [1])

So what should it be? A literal reference to Cleveland's n_(s), n_(l), and
n_(t) would be really helpful here.

Thank you,

[1] https://www.wessa.net/download/stl.pdf
[2] well, technically he does in step 2 of the algorithm, but it is not
called a low-pass filter in the paper.

	[[alternative HTML version deleted]]

More information about the R-devel mailing list