See attached.
Hi Maria,
After the lecture and discussion in class last week, I agree that using the clinic as it's own control will be the most straightforward analysis, and that measuring multiple time points to create an interrupted time series will add analytic integrity beyond a simple before/after study. I plan to switch my study design to an interrupted time series as well.
One clarification I realized that I need, however, is that it seems like the chi-squared statistic would be the most appropriate for both our studies, since we will be comparing proportions of dichotomous variables, yet the example statistics presented in lecture were t-tests. Perhaps someone else in group has the answer?
Maria O.