Pearson's chi-test for more than 2x2

[Trolle]

So, I have one stat book (Discovering Statistics by Andy Field; great book) that discusses Pearson's chi-test and how it's only applicable to a 2x2 contingency table with categorical data. The solution is to use the loglinear analysis for larger tables according to Andy.

Another book I have (Statistics for Economics, Accounting and Business Studies by Mike Barrow; another easy-to-read stat book) does not specify that a chi-test is restricted to a 2x2 table and has an example with more than 2 variables.

So, which one should I put my faith in? I guess it should be said that Andy's book, although written more or less for a general audience, is more geared towards experiments while Mike's book is more business-oriented. My field is also business, if that makes any difference.

I am asking for my own curiosity and that I can't do loglinear analysis on my version of SPSS since the school hasn't doled out for the fancy add-ons. I did see this post but the answer a link to a Wiki page that basically said chi-square is normally used for 2x2 but also larger tables.

 

[noetsi]

Chi square is not restricted to a two by two table. It is commonly run with more than two levels of a variable. When it is a two by two table it reduces to a z test of preportions (I have also heard a fischer test although that has different assumptions) and that might be what the original book was saying. The variables dont actually have to be categorical (nominal) in nature but chi square will treat them like they were.

Really strange that an author would say chi square has to be two by two.:shakehead

 

[Trolle]

Thanks for the response. Fisher's exact text is also mentioned in the reading but it is repeated several places that a 2x2 table is the way to go.

Glad to know that I can expand that now.

Thanks.