Re: chi-squared test. hypothesis confusion
- From: Bruce Weaver <bweaver@xxxxxxxxxxxx>
- Date: Fri, 23 Dec 2005 19:29:35 -0500
eblabac@xxxxxxxxxxxxxxxxxxx wrote:
This is a common question when starting to do hypothesis testing. Your tests H0 is smoking and allergy are statistically independant, and Ha is smoking and allergy are not independant. From your notes you have a decision rule associated with this chisq test. You reject H0 if your test statistic chi^2 >= chi^2 alpha (where chi^2 alpha, (r-1)(k-1) is found in a table or generated from a program depending on your level of alpha and r and k). Simply find the chi^2 alpha value from a table (in any stat book), then calculate the chi^2 statistic using your data. Then follow the rule ... if chi^2 >= chi^2 alpha - reject Ho, i.e. smoking and allergy are NOT independant, if chi^2 < chi^2 alpha you dont accept Ha, you "fail to reject H0", i.e. smoking and allergy are independant. Its as easy as that. However, make sure your data satisfies any assumptions of the test. This particular test has very flexible assumptions, in that you only need your observations to be independant of each other and that each cell of your 2x2 table is greater than 0 (preferably bigger than 5, otherwise you could get skewed results). Hope that helps.
Eric B.
It is the expected frequencies, not observed, that need to be equal to 5 or more for a 2x2 chi-square. Here is a summary of the conditions under which chi-square tests perform reasonably well.
www.angelfire.com/wv/bwhomedir/notes/chi_square_assumptions.txt
-- Bruce Weaver bweaver@xxxxxxxxxxxx www.angelfire.com/wv/bwhomedir .
- References:
- chi-squared test. hypothesis confusion
- From: mikeeria
- Re: chi-squared test. hypothesis confusion
- From: eblabac
- chi-squared test. hypothesis confusion
- Prev by Date: Re: chi-squared test. hypothesis confusion
- Next by Date: Re: chi-squared test. hypothesis confusion
- Previous by thread: Re: chi-squared test. hypothesis confusion
- Next by thread: Re: chi-squared test. hypothesis confusion
- Index(es):
Relevant Pages
|