An application of methods for clustered binary responses to a cardiovascular study with small sample size.

This paper discusses statistical methods for a cardiovascular study in which each of eight animals had a dichotomous outcome observed for each of several treatments. There were five treatments in all: shunt, control, two doses of a test drug for potentially causing an unfavorable cardiovascular event, and a combination of the test drug and a counteracting agent. Exact conditional methods were used through LogXact, a statistical software for exact logistic regression and an alternative framework for performing a large class of nonparametric tests performed by StatXact. The results agreed reasonably with asymptotic methods even though the sample size was small.