Re: Chi-Square

Marc Weeber (M.Weeber@farm.rug.nl)
Wed, 12 Mar 1997 11:15:36 +0100 (CET)

On Wed, 12 Mar 1997, Michael Klotz - englische Sprachwissenschaf wrote:

Hi Michael,

your question has been discussed earlier on this list.

for an overview of this discussion, see the collection made by
Tony Berber Sardinha:

http://www.liv.ac.uk/~tony1/compacor.txt

Adam Kilgarriff has discussed this topic you're interested in extensively.
For an overview: see

http://www.itri.brighton.ac.uk/~Adam.Kilgarriff/aisb.ps.gz

See for more references, this paper I wrote:

http://www.farm.rug.nl/marc/artikel.ps.gz

conclusions are in short: don't use CHI-square, but do use the
log-likelihood ratio. See also Ted Dunnings article in Computational
Linguistics:

@Article{dunning:1993,
author = {Ted Dunning},
title = {Accurate Methods for the Statistics of Surprise and
Coincidence},
journal = {Computational Linguistics},
year = 1993,
volume = 19,
number = 1,
pages = {61--74}

happy comparing :-)

Marc

> Dear List Members,
>
> I am currently working on the question of whether there is a
> correlation between certain grammatical choices concerning English
> verbs and the lexical environment they are in. Let me briefly
> illustrate this: it can be shown from corpus data that the verb
> 'bring' in collocation with the noun 'charges' as object occurs in
> the passive far more often than it does on average without looking
> at any particular collocation.
>
> In order to show correlations of this type on a statistically sound
> basis I thought that the chi-square test might be appropriate. I
> would be grateful for any hints as to the appropriateness of this
> test for my purposes. Any information about literature describing
> the use of the chi-square test in connection with lexical
> distributions would be most welcome, as well.
>
> thanks
> Mike
> _________________________________________________
> Michael Klotz (mklotz@phil.uni-erlangen.de)
> Institut f. Anglistik und Amerikanistik
> Lehrstuhl f. engl. Sprachwissenschaft
> Friedrich-Alexander-Universitaet Erlangen-Nuernberg
> Bismarckstr. 1
> 91054 Erlangen
> GERMANY
> Tel: 09131-852938
>

------------------------------------------------------------------
Marc Weeber http://www.farm.rug.nl/marc/home.html
Groningen University Centre for Pharmacy
marc@farm.rug.nl Social Pharmacy and Pharmacoepidemiology
tel: +31 50 3637571 ___ A. Deusinglaan 2
fax: +31 50 3633311 | 9713 AW Groningen, The Netherlands
----------------------0-------------------------------------------