Monitoring and Evaluation of Smolt Migration in the Columbia Basin Volume VIII : Comparison of the RPA Testing Rules, Technical Report 2002.

The 2000 FCRPS Biological Opinion (BO) suggested two statistical hypothesis tests to assess the RPA compliance by the years 2005 and 2008. With the decision rules proposed in the BO, Skalski and Ngouenet (2001) developed a compliance framework based on classical t-tests and used Monte-Carlo simulations to calculate power curves. Unfortunately, the two-sample t tests proposed in the BO only have moderate-to-low probability of correctly assessing the true status of the smolt survival recovery. We have developed a superior two-phase regression statistical model for testing the RPA compliance. The two-phase regression model improves the statistical power over the standard two-sample t-tests. In addition, the two-phase regression model has a higher probability of correctly assessing the true status of the smolt survival recovery. These classical statistical power curve approaches do not incorporate prior knowledge into the decision process. Therefore, we propose to examine Bayesian methods that complement classical statistics in situations where uncertainty must be taken into account. The Bayesian analysis will incorporate scientific/biological knowledge/expertise to thoroughly assess the RPA compliance in 2005 and 2008