Model checking is an important part of statistical practice and exploratory data analysis. When teaching a first class in regression, for example, we advise that students plot residuals to check that the regression model is capturing the features evident in the data. In Bayesian inference, predictive checks have been recommended as a way to examine models (Rubin, 1984; Gelman et al, 1996). For exploratory data analysis, these checks can provide new insight or suggest improved models for the process that generated the data.
However, a stopping rule can affect predictive checks, and cases in which the stopping rule is not completely specified can be particularly challenging. We present an example in which the task is to design a wildlife survey using existing data subject to a partially-known stopping rule. The example illustrates the value of predictive checks even in cases where the models under consideration are complicated, and the effect of the stopping rule might be considered to be ignorable.