Date: Thu, 17 Feb 2005 20:05:37 +0000
Reply-To: Rani Vohra <firstname.lastname@example.org>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: Rani Vohra <email@example.com>
Subject: Goodness of Fit Test Apply? Two-fold question:
Content-Type: text/plain; format=flowed
I am trying to determine which tests apply to my data, and I am a bit
I have collected data (1000's) about how well an industry test accurately
passes a production item.
The data looks something like:
After testing an item once - an item has a 40% chance of passing the test
After testing an item twice (with the same test!) - an item has a 60% chance
After testing a item three times - at item has a 70% chance of passing
And so forth, until about 5 re-tests result in about 95% rate of passing.
Obviously, there is something wrong with the process and I will be
attempting to narrow down the causes.
I have noticed that the standard deviations also decrease with each
re-test. Does the F-test accurately apply here to determine if they are
statistcally significant? And if so what, what does that suggest? Human or
My second question is once I have implemented some change and have collected
new data for the same categories, does it make sense to use the Chi-Squared
Goodness of Fit test to determine if there is a statiscally significant
change, or the Wilcoxon Matched-Pair Signed Test of Differences...
I think (though I'm not sure) that a non-parametric solution is best as the
data starts with it's highest peak, and then tapers off.
I would really appreciate any suggestions. THANKS!