
Selecting Distributions 121
• Approach 1. Using p = 1/50 = 2% and thus ignoring the uncertainty
we have regarding the results of the rst study
• Approach 2. Using the Beta distribution to represent the uncertainty
around p
Figure4.24 shows how to construct both alternative models (with two ver-
sions of the second method: the construction approach, Approach 2, and the
simpler Beta distribution, Approach 3) to simulate the number of side effects
in the second study.
Approach 1 simply uses the Binomial distribution and assumes that the
observed side effect rate (.02) from the small study is the actual side effect
rate, known with certainty. A ...