Project Summary/Abstract In the early stages of research pilot data is often collected to determine if the research plan for a follow-up larger trial is feasible; however, determining if the research hypothesis itself is plausible after collecting the initial data is a difficult problem. Traditional hypothesis testing is poorly suited to this task due to the disconnect between the goal of hypothesis tests, to provide confirmatory evidence on the endpoints of interest, and the need for the pilot study analysis, to determine if a follow-up trial is justified. Other approaches have been proposed but these often require knowledge of a minimum clinically important difference (MCID), in novel research an MCID is not typically known. Researcher's need a way to determine if their theory is being borne out based only on the observed pilot data. Additionally, these early analyses are often complicated by the collection of multiple endpoints of interest. We've developed a global hypothesis test, the prediction test, which is intended for use with many endpoints of interest relative to the sample size, allows for the endpoints to be correlated, controls the family-wise error rate, and increases in power as the number of endpoints increases. We propose extensions to the methodology of the prediction test that will make it more powerful and allow it to be used in more settings. Specifically, we will extend the methodology to allow for more types of predictions, allow the prioritization of endpoints, account for the magnitude of observed effects and increase overall power by addressing the conservative nature of the test when the number of endpoints is very small. Using these methodological developments, we will perform a secondary data analysis on two Alzheimer's Disease pilot studies. This will enable us to re-examine the original hypotheses which dealt with the effect of aerobic exercise on Alzheimer's outcomes, and the potential of a dose-response for Aerobic exercise in pre-clinical Alzheimer's Disease patients. The prediction test can better address the original research hypotheses by combining the information across all endpoints, both primary, secondary, and exploratory. Additionally, the prediction test is well suited to the evaluation of small but consistent effects across many endpoints and addresses a more appropriate hypothesis than the traditional change in central tendency, namely, we address the hypothesis of whether the researcher's theory is predictive of the data. Thus, the re-analysis of these studies using a more powerful and appropriate test will provide better understanding of the effects of aerobic exercise on the overall health (cognitive, physical and function) of both AD and pre-clinical AD patients.