|
Abstract:
|
We will discuss some practical considerations when using a robust mixture of conjugate priors. As indicated by Schmidli et al., (2014), a non-informative conjugate component of a mixture prior allows for faster discounting when prior-data conflict exists. However, their work focused on the meta analytic predictive prior (MAP) as the informative conjugate part of the robust mixture prior. Weaver et al., (2016) assessed two mixture priors; the robust MAP and the simple robust conjugate prior for assessing adverse events. Their work seems to suggest that a simple robust conjugate prior may be preferable over the robust MAP in trials with small sample sizes, as sample size may play a role in optimizing this adaptive discounting feature when prior-data conflict exists. We will present and discuss results from our investigation into this possible asymptotic characteristic through simulations using the safety example presented by Weaver et al., (2016). Simulations will involve a single sample assessment which is common for non-interventional type trials, as well as absolute and relative group assessments of active versus placebo which is common for randomized controlled trials.
|
