Re: Chi-Square

Chris Brew (
Wed, 12 Mar 1997 14:09:13 +0000

>In relation to Chris Brew's comments, he covers one problem with
>chi-square, but these are further problems.
Quite. Different statistics do different jobs. My earlier message
was intended to indicate a situation in which chi-square is very
likely to mislead. Adam's paper gives another. I certainly didn't
mean to suggest that chi-square is ever the method of choice.
The essential point is that corpus linguists need to be statisticians,
at least to some extent, if they are to avoid getting in a tangle.