A stats question for all you brainiacs....please help!?

Question

I am working on a project and it has been a while since I took stats. I am comparing drop out and retention rates among a particular group of students, before and after the implementation of new lesson plans/curriculum, to determine if the new curriculum made a difference. I'm supposed to use a two-tailed test since I don't know which direction the results will go in, right? (Obvioulsy I don't know for sure!) Also would I be using a z-test, t-test, chi square! I just don't remember anything about this and it's not my expertise!

So the null hypothesis is that implementation fo the new curriculum will not make a difference in the two groups. I will not reject the null hypothesis unless the difference in favor of the new curriculum is so big that it would have only a small chance due to sampling variation. Is this right? If so I guess I should use a one-tailed test???

Should I be basing this comparison on mean? I'm totally confused...

Answer 1

Since you're comparing rates (the percentage of drop-outs in the two groups), most likely you'd do a 2-proportion z-test for your project. You're not looking at mean, but rather proportions (# who dropped out or stayed out of the total).

As to whether you whether you do a two-tail test, it depends on exactly how you phrase your hypotheses. Do you want to know whether the new curriculum improves the retention rate or whether the retention rate is different with the new curriculum (could be higher or lower)? The first would be a one-tail test; the second two-tail. What you said here ("will not make a difference") is indeed a 2-tail test.