see attached
Hi Maria,
I think your choice of using the clinic as the unit of analysis is a good choice given this would avoid contamination of the control subjects. However, given that you work with residents, I'm wondering whether these residents will go to your control clinics, as well? Could you further discuss why you would choose a step wedged approach over the cluster randomization we discussed in class? I'm just thinking about the concerns brought up in class about the step-wedged approach being challenging from a feasibility standpoint.
Thanks,
Vicky
Hi Maria,
You bring up a good poing about chi-squared tests vs t-tests. I think the answer may be in Chuck's slides, on slide 16, in which he says that ch-square tests are wrong for this type of analysis because they assume that everyone is independent, and so t-tests are they test of choice. Though there may be other differences in the analysis that may be accounting for the choice of tests. Let's see if others have a good answer!
Maria