Accelerated life testing is widely used to quickly obtain lifetime data of a product. On the other hand, degradation testing can be also used to estimate the lifetime of a product by relating performance degradation data with lifetimes. This testing has an advantage that the lifetime of a product can be predicted before any specimen fails. Accelerated degradation testing employs accelerationg stresses that hasten the degradation of product performance, and this may further accelerate the test. In this thesis, we develop statistically optimal accelerated degradation test plans under the assumption of Arrhenius degradation rate. As a criterion of optimization, we adopt the asymptotic variance of the estimator of the mean of the lifetime distribution. Also, such optimal accelerated degradation test plans are compared with optimal accelerated life test plans. The comparison results show that accelerated degradation test plans are in general better than accelerated life test plans unless the probability of failure at time 0 is large.