Re: Significance testing of Dependent proportions
- From: Bruce Weaver <bweaver@xxxxxxxxxxxx>
- Date: Fri, 16 Sep 2005 06:48:08 -0400
Hari wrote:
Hi Richard,
Thanks for the response.
I goggled for Mcnemar and got this link http://www2.chass.ncsu.edu/garson/pa765/mcnemar.htm and http://www.public.asu.edu/~huanliu/dmml_presentation/T-test.pdf
After going through the text and the example in the above 2 links am sure this is not what I want as my data structure is different. (I dont have BEFORE and AFTER scenario)
Let's say I have a universal set consisting of 100 people who ate at Shop A or Shop B or Shop C or Shop D during today's lunch. (everybody ate at one and only one of these 4 shops).
Let's suppose I construct a set of people who did not eat at shop A (equal to 70).
Now, everybody who has not eaten at shop A would have eaten at shop B, C or D. Let's say these numbers are 20, 40 and 10. (20 + 40 + 10 = 70)
Now, I want to know whether there is any significant difference between number of people who did not eat at A with number of people who ate at B (and afterwards number of people eating at C and D).
Please suggest the formula or test for the same. (accordingly I can then code it out in excel).
Regards, Hari India
I still don't understand your question. Are you asking how to partition the overall (goodness of fit) chi-square into orthogonal components?
-- Bruce Weaver bweaver@xxxxxxxxxxxx www.angelfire.com/wv/bwhomedir .
- Follow-Ups:
- References:
- Significance testing of Dependent proportions
- From: Hari
- Re: Significance testing of Dependent proportions
- From: Richard Ulrich
- Re: Significance testing of Dependent proportions
- From: Hari
- Significance testing of Dependent proportions
- Prev by Date: Re: Significance testing of Dependent proportions
- Next by Date: Anything better than pairwise correlation testing
- Previous by thread: Re: Significance testing of Dependent proportions
- Next by thread: Re: Significance testing of Dependent proportions
- Index(es):
Relevant Pages
|