Value of Information for Comparing Dependent Repairable Assemblies and Systems A designer can measure reliability using various metrics such as the probability of failure of a component or system and the failure rate. The choice of the metric depends on the type of the component or system under consideration. Making a reliability-based decision between design alternatives will depend on metrics such as the failure probability or the failure rate. This paper presents an approach for optimizing sample sizes for a test so as to maximize the expected utility of a decision that involves dependent design alternatives based on the repair frequency. Such dependency means that as a designer collects information on one design, the decision maker begins to infer the reliabilities of the other designs. The approach uses principals of decision theory to determine the sample size so as to maximize the expected utility of a decision. The approach uses a Bayesian probability model, which allows the decision maker to incorporate subjective priors on the reliability performance of the design alternatives. The dependency is modeled using Copulas to “couple” the marginal prior distributions of the alternatives to a single, joint prior. The paper uses Markov Chain Monte Carlo simulation (MCMC) to determine the posterior probability density and the resulting expected utility of the decision. The paper will evaluate design alternatives based on their failure per unit (FPU) and by considering Archimedean Copulas to couple the dependent marginals that describe the priors for each design alternative’s failure per unit behavior. This paper is an extension the paper "Assessing the Value of Information for Multiple, Correlated Design Alternatives" (Capser and Nikolaidis, 2017), which presented an approach for determining optimal sample sizes for assessing correlated non-repairable design alternatives based on the prior estimate of their joint failure probability.