Hi, students –
Some items you may find helpful for HW #3:
1. Sample size calculator clarifications:
A.) T-test for continuous predictor and dichotomous outcome: Let's suppose you are doing a study to see whether a measure of maternal stress (a continuous predictor variable) predicts preterm birth (a dichotomous outcome). We can use a t-test for this, but we need to turn things around from what the sample-size calculator is expecting, because in this case the group variable is the outcome – preterm birth or not – not the exposure. So if we use the Means-Sample size calculator at sample-size.net, the number we would enter for "q1 Proportion of subjects that are in Group 1 (exposed)" would actually be the proportion of the mothers expected to deliver a preterm infant (e.g., 8%). Similarly, if we used the Means-Effect size calculator, the numbers we would enter for the group sizes (N1 and N0) would be the numbers with and without a preterm birth (and the ratio of N1 to N0 would be 8:92).
B.) Continuity correction: As Warren said in class, whether or not you use the continuity corrected sample size estimate does not make much difference unless your sample size is small. But I would go farther and say the continuity corrected estimate is the one you should always use. The sample size you calculate otherwise will depend on you using the less accurate uncorrected test to analyze the data, which we do not recommend.
So why does the website include sample size estimates we don't recommend you use? The main reason is so if you use some other calculator and get a different answer, you will be able to see if the reason why the answers are different is because that other calculator does not use the continuity correction.
2. Homework uploading
Some have asked about uploading homework to the CLE website. It is possible that having everyone do this might save section leaders some time, but I am not convinced of this, so I am leaving it up to section leaders. Unless you hear otherwise from your section leader, you can continue to submit your homework assignments via e-mail.
Feel free to post any questions to the forum and your teachers and fellow students will do their best to answer them!
I hope you are enjoying the course and learning a lot and not too traumatized by your sample size calculations!
Yours,
Tom