A Multi-scale Bayesian Hypothesis Testing Approach for the Analysis of Prepulse Inhibition Test

Prepulse inhibition (PPI) deficits have been seen in various neuropsychiatric disorders, such as schizophrenia, Tourette’s syndrome and Huntington’s disease. How to effectively distinct these PPI deficit subjects from normal ones is a challenging task because the PPI data measured from animal experiments are usually erroneous, incomplete, and unstable. In this paper, we introduce a novel multi-scale Bayesian hypothesis testing (MBHT) method to rectify the PPI data. Specifically, we regard any startle response signal as a deteriorated result from a hypothesis and reconstruct the complete and stable signal using a Bayesian hypothesis testing approach. The multi-scale analysis here is necessary because we cannot directly estimate the proper scale for a hypothesis even we have some statistical quantities of the whole population. By carrying out two sets of animal experiments using different medicines, we show that this MBHT method is much more robust than the conventional method. The analytic methodology proposed in this work can also be applied to similar animal experiments.