In my field I see lots of papers constructing dose response curves for various drugs in mouse/rat disease models. They usually do a single test at each drug dose and have a quantal (yes/no) measurement of some kind of disease characteristic. They say they are able to test whether the curves are significantly different using tests like t-test or Fisher exact test, specifically the ED50s. The values at each dose are a single percentage value - e.g. % responders in the test group. I don't understand how these tests can be performed on data like this and can't find an explanation of how to do it online. Can someone explain it to me? Instructions for Graphpad Prism only shows how to do it if there are 3 replications for each dose. Doesn't seem to be possible for just one.