Skip to content
Prev 35811 / 63424 Next

Incorrect Kendall's tau for ordered variables (PR#14207)

Dear Peter,

Thank you. Although the documentation does mention numeric variables,
one would intuitively expect cor() and cor.test() to work for ordered
factors with methods "kendall" and "spearman". After all, these are
nonparametric procedures, defined for ordinal scales, and the only
information they need are ranks (the same should be true for
wilcox.test()).

So even if this is, strictly speaking, not a bug I would strongly
suggest extending cor() and cor.test() to work with ordered factors for
Kendall's and Spearman's correlations (although this would not make much
sense for Pearson's correlation). It looks like the change should be
very easy.

Marek Ancukiewicz
-------------- next part --------------


The information in this e-mail is intended only for the ...{{dropped:13}}
Message-ID: <20100208140701.0B5EE5E1B0@biostat.mgh.harvard.edu>
In-Reply-To: <4B70103C.7030504@biostat.ku.dk> (message from Peter Dalgaard on Mon, 08 Feb 2010 14:23:08 +0100)