You would probably want to transform the proportion using, .e.g. the arcsine transformation or the logit.
I spoke too quickly about t-test, which, you are correct, wouldn't be right with raw proportions.
But, as I noted, the bigger issue is what the data mean, not whether they are significant.
6.9% vs. 6.4% is, for most applications, a very small difference. Whether this is so for you is something you would know.
>>> Larry Kaskey <LKaskey@TRACFONE.COM> 04/01/02 01:54PM >>>
Thanks Peter, but I thought, perhaps wrongly, that a t-test would work for
continuous data, such as comparing the time it takes for a person to redeem
the coupon, where a chi square would be better for catagorical data such as
in my case, response or no response.
My statistics is very rusty so I am easily influenced!
Thanks for the help,
From: Peter Flom [mailto:flom@NDRI.ORG]
Sent: Monday, April 01, 2002 1:43 PM
Subject: Re: Chi Square interpretation
One question is why do Chi-sqare rather than a t-test?
You have two samples, and a proportion from each who did something.
And you can then get a confidence interval around the difference.
However, on the data you gave, it seems to me like your conclusion is
correct. It just might be a type 3 error - solving the wrong problem
The other big question is, whether or not any test is significant or not,
you care, on a substantive level, about the difference in the two rates.
Even if 6.9% and 6.4% are exactly the right values, would you want to go
Peter L. Flom, PhD
Assistant Director, Statistics and Data Analysis Core
Center for Drug Use and HIV Research
National Development and Research Institutes
71 W. 23rd St
New York, NY 10010
(212) 845-4485 (voice)
(917) 438-0894 (fax)