Multivariate Autoregressive State-Space Modeling
Description
The MARSS package provides maximum-likelihood parameter
estimation for constrained and unconstrained linear
multivariate autoregressive state-space (MARSS) models fit to
multivariate time-series data. Fitting is primarily via an
Expectation-Maximization (EM) algorithm, although fitting via
the BFGS algorithm (using the optim function) is also provided.
MARSS models are a class of dynamic linear model (DLM) and
vector autoregressive model (VAR) model. Functions are
provided for parametric and innovations bootstrapping, Kalman
filtering and smoothing, bootstrap model selection criteria
(AICb), confidences intervals via the Hessian approximation and
via bootstrapping and calculation of auxiliary residuals for
detecting outliers and shocks. The user guide shows examples
of using MARSS for parameter estimation for a variety of
applications, model selection, dynamic factor analysis, outlier
and shock detection, and addition of covariates. Type
RShowDoc("UserGuide", package="MARSS") at the R command line to
open the MARSS user guide. Online workshops (lecture material
and labs) at
http://faculty.washington.edu/eeholmes/workshops.shtml