[Corpora-List] effect size

[Tina Waldman]

Dear List members


I want to calculate the effect size for my chi-square stat. findings in a comparison of the number of collocations occurring in 4 corpora standardized to 40,000 words. A significant difference was found between the 4 groups   (Chi-square (df3) = 73.39, p<.0001). Additional comparisons between each pair of groups revealed  significant differences between corpus 1 and 2(chi-square = 180.18, p < 0.001), corpus 1 and 3 (chi-square = 74.93, p < 0.001) and corpus 1 and 4 (chi-square 55.77, p < 0.001). 

I read that effect size is done by computing the square-root by dividing the sample size. Does this mean I divide the sample size of the running words in the corpus? Do I use the raw frequencies? Do I divide the number of collocations? 

Would be pleased of any help

Yours

Tina Waldman

 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/corpora/attachments/20090228/c4e6abfc/attachment.htm>
-------------- next part --------------
_______________________________________________
Corpora mailing list
Corpora at uib.no
http://mailman.uib.no/listinfo/corpora