Bayesian structural TS

[BinhPham]

Dear Tom,

I have a look at Predicting the Present with Bayesian Structural Time Series by Scott and Varian (2013) and Scott' R package bsts, and I do think it is so close to SSM (with DLM) but the MCMC estimation approach.

Could we have the similar thing with DLM and RATS Bayesian facility? I understand the paper and SSM in RATS but lacking experience to combine things to work. Just need an advice for playing around as I am curious on how we can estimate SSM parameters using Bayesian methods with RATS.

This link http://multithreaded.stitchfix.com/blog ... get-arima/ (Sorry ARIMA, but I’m Going Bayesian) also provide a nice application of above package.

Thank and best regards,

[TomDoan]

There are quite a few examples of Bayesian estimation of state-space models. The koop_bp191.rpf is an example of a structural time series model. The key instruction is DLM with TYPE=CSIMULATE which does a conditional (on the observed data) simulation of the states and the shocks. Given a conditional simulation, the shocks are treated as data and are drawn using straightforward sampling methods.