The first part of this thesis addresses the challenge of efficiently detecting changes within a network of sensors, where minimizing communication between sensors and the cloud. We proposed two online, communication-efficient methods to detect such changes. We provide an asymptotic theory for the first method, disMOSUM, concerning consistency and the asymptotic distribution if there are no changes. Simulation results suggest that our method can achieve similar performance to the idealised setting, where we have no constraints on communication between sensors, but substantially reduce the transmission costs. The second approach, mixFOCuS, addresses the scenario where post-change parameters are unknown and the data belong to the exponential family, while still maintaining computational efficiency. A simulation study is conducted to evaluate the performance of our method with state-of-the-art approaches.
In the second part, we consider Bayesian approaches to online changepoint detection in linear regression models. Such methods are less common than frequentist methods due to their perceived higher computational cost, but they have advantages in terms of more naturally quantifying uncertainty and the ability to incorporate prior information about the type of change. We proposed a fast online Bayesian changepoint algorithm that can be applied to a wide range of problems, including detecting changes in mean or slope, and detecting changes in the presence of seasonal effects. Simulation suggests that the algorithm has a similar speed but higher accuracy compared to a benchmark pruning approach, which only prunes the candidate with the lowest posterior probability.