L-Ippel/SEMA: Streaming Expectation Maximization Approximation

The SEMA algorithm fits multilevel models in the context of streaming data (a.k.a. data streams). Using online learning, the model is updated a data point at the time. SEMA algorithm only stores summarizing statistics, which limits memory usage, and because it never returns to the data, sema algorithm is computationally efficient. The current state of the algorithm is able to fit multilevel models with fixed and random effects.

install.packages("remotes")
remotes::install_github("L-Ippel/SEMA")