Question

Home > Statistics Every Writer Should Know > The Stats Board > Discusssion

Question
Message posted by Tom (via 64.12.103.182) on June 9, 2001 at 6:19 PM (ET)

Campus security states that only 5% of the cars parked on campus are parked illegally. A random sample of 250 cars parked on campus showed that 15 cars were parked illegally. Test the hypothesis at level .05 that the percentage is too low.

??is this a two tailed test??

??? Ho:u1-u2=15 vs Ha:u1-u2not= 15

what is the proper way to set this up????

READERS RESPOND:
(In chronological order. Most recent at the bottom.)

Re: Question
Message posted by JG (via 128.8.22.76) on June 9, 2001 at 8:35 PM (ET)

This is a one-sided test. The p-value is .44, so that at the .o5 level we accept the hypothesis that there is no evidence that 5% is too low.

Re: Question
Message posted by Stat (via 209.81.243.198) on June 14, 2001 at 10:20 AM (ET)

One-tailed t-tests should rarely (if ever) be used when performing statistical tests. I'll explain why below. First, however, your hypotheses are set up incorrectly. That is how you would do a 2-sample t-test. You are comparing one sample, p (15/250), to a constant (5%). So your null hypothesis should be that p = .05.

When you select an alpha value (.05 in your case), you are saying that you want the chance of committing a type I error (rejecting the null hypothesis when it is actually true, meaning you say that campus security is wrong when it was actually right) to be <=.05. When you do a 1-sample t-test, it becomes easier to commit a type I error on the side you choose (in this case it would be p>5%, and easier to commit a type II error on the other side (p<5%) because your null hypothesis rejection zone is all on the > side. That is why you should (almost) always use a 2-tailed t-test.

When you do that in this case, your Ha becomes p not= 5%. The P-value becomes .47, so you cannot reject Ho; you can't prove that campus security was incorrect.

You were right, and JG was wrong.

Your $5 contribution helps cover part the $500 annual cost of keeping this site online.