Please help me with this statistics question?

Question

. If a researcher were studying the effects of a teaching method on patient learning outcomes, how would the research question need to be worded to use a t-test to test for statistical differences? What type of data would need to be collected and why? Thank you

Answer 1

Please also refer to the answer to your other question:
http://answers.yahoo.com/question/index;_ylt=AijQ0Q62UlqXbrNXY7DW7j3sy6IX;_ylv=3?qid=20071221122239AAVa3Wv&show=7#profile-info-594bfadade4b265031cda0824b9cf842aa

The only way to use a parametric test would be to use a standardized test to score for the multiple teaching methods.

If you have a standardized test then you can use a t-test for the difference in means between two teaching methods.

if you have ordinal data then you need to use a chi-square test.

You also have to consider other possible factors in the outcomes such as the student's age, gender, the semester the class is held, the type of institution.... etc. With multiple factors use a general linear model and ANOVA for the analysis. This will help you determine if the teaching method is a factor in the students' performance or if something else was influencing their performance.

Answer 2

The research question could NOT be a yes/no question. The answer would need to result in some kind of continuous numerical value.

For example, you cannot ask: Do you understand the lesson?
You would wind up with a dichotomous result, which cannot be analyzed with a t-test.

But you can ask: On a scale of 1 to 5, how well do you understand the lesson?
Then you'd have a relatively continuous numerical result that can be averaged and comapred with a t-test.

As for Dave's analysis of t versus z tests, I think that's too simplistic. The z-test is intended to compare whole populations, assuming normal distribution. In the real world, if we wanted to do z-tests, we'd rely on census data collected from thousands of people. In the context of a clinical trial, we would have dozens or hundreds of test subjects, but we would still use the t-test.

Answer 3

t-test = a test with # of observations under 30,
z-test = a test with # of observations 30 and over.