A scalable quasi-Newton estimation algorithm for dynamic generalised linear models

This research develops a scalable computing method based on quasi-Newton algorithm for several estimation problems in dynamic generalised linear models (DGLMs). The new is developed by applying the principle of maximising pointwise penalised quasi-likelihood (PPQ) DGLM observed data often massive size. Statistical and computational challenges involved this development have been effectively tackled exploiting specific block structure sparsity underlying projection matrix. obtained maximum PPQ estimator state vector has shown to be consistent asymptotically normal under regularity conditions. Numerical studies real applications are conducted assess performance method.