MessyTimeSeries.jl
MessyTimeSeries.jl copied to clipboard
State space models with diffuse initial conditions
Add a simple implementation to handle state space models with diffuse initial conditions.
@azev77 I have implemented the so-called sequential / univariate approach to Kalman filtering and smoothing. This approach is used instead of the default filtering / smoothing by default if the covariance of the measurement errors is a UniformScaling{Float64}. I am planning to extend it to handle the exact diffuse initialisation in Durbin and Koopman (2000).