dajmcdon/spasm: Implements the Laplace-Gaussian filter for sparse, additive, state-space models

Estimates semi-parametric high-dimensional state-space models. Here we presume that both the observation and state distributions have Gaussian noise, but we let the observation equation evolve as an additive model in the state. In particular, we expand the state in the spline basis and estimate the coefficients via the group lasso. As this model is non-linear, we use the Laplace-Gaussian filter to estimate the hidden states quickly and accurately.

install.packages("remotes")
remotes::install_github("dajmcdon/spasm")