Home > Statistics Every Writer Should Know > The Stats Board > Discusssion
Which stat to use?
Message posted by Terry on November 22, 1999 at 12:00 AM (ET)
I'm taking a research class and had to conduct an experiment. I have one group of 10 people and provided them with a choice to drink bottled water or city water. 5 chose bottled and 5 chose city. I then showed them bottled water ads. 8 chose bottled water and 2 chose city after the ads. Would I use the Chi Square Independence Test to test my hypothesis? I tried the pre-test post-test on same group--dependent t test but it doesn't look right. I also wondered about Chi Square Goodness of Fit. Any help would be appreciated. My paper is due soon! Thanks!
READERS RESPOND:
(In chronological order. Most recent at the bottom.)
Re: Which stat to use?
Message posted by JG on November 23, 1999 at 12:00 AM (ET)
Try chi-squared test for independence. Remember, you have very little data so even an appropriate test is going to be 'weak' .
Re: Which stat to use?
Message posted by Bill on November 23, 1999 at 12:00 AM (ET)
First, what is your hypothesis?? Without knowing it is difficult to answer your question. If your hypothesis is that subjects' use of bottled water at Time 2 is independent of their use of bottled water at Time 1 then you could use a "repeated measure" chi-square. If your hypothesis is that the number of changers (from bottled water to non-bottled water) does not differ from the number of changers in the other direction (non-bottled water to bottled water) then McNemar's chi-square is appropriate. The following applies to both tests: As you know, each observation in a chi-square must be represented only once in the 2x2 table. Thus, you must classify each subject as a bottle water user (BW) or a non-bottled water user (NBW) at times 1 and 2. Each subject would be a BW-BW, BW-NBW, NBW-NBW, or NBW-BW. For example, a non-user at time 1 and user at time 2 would be a NBW-BW. The following table may help. (Obviously, you need to know more than the 5/5 and 8/2 split that you described. You must know how each subject responded at time 1 and at time 2.) Having recorded the data as described, the repeated measures chi-square is computed as a regular 2x2 chi-square. The McNemar uses a different formula. Let me know if the above is not clear enough. ---------------------------------Time 1
-----------------------------BW NBW
Time 2---------BW
------------------NBW
Your $5 contribution helps cover part the $500 annual cost of keeping this site online.