Experimental design and model selection: The example of exoplanet detection

Abstract

We apply the Minimum Description Length model selection approach to the detection of extra-solar planets, and use this example to show how specification of the experimental design affects the prior distribution on the model parameter space and hence the posterior likelihood which, in turn, determines which model is regarded as most `correct'. Our analysis shows how conditioning on the experimental design can render a non-compact parameter space effectively compact, so that the MDL model selection problem becomes well-defined.