Customer Reviews

Bayesian Forecasting and Dynamic Models (Springer Series in Statistics)

5 star:		(3)
4 star:		(0)
3 star:		(0)
2 star:		(0)
1 star:		(0)

 

 

A Bayesian approach is a natural way to deal with time series data. You construct a model based on past data and prior information and use the model to predict future values in the series. When the new observations come in the model can be updated (model parameters reestimated) and forecasts can be updated. Most of the time series literature deals with the classical (frequentist) approach incluing the well-known book by Box and Jenkins on forecasting and control. This book provides a mathematically rigorous treament of time series modeling based on a Bayesian approach. Many common forecasting procedures including the Kalman filter are iterative algorithms that could be derived as solutions for forecasting based on a Bayesian model of the time series.

This is the best text available on this topic.

As a reader with an economical background, mathematical texts are usually hard to be followed. Nevertheless, dinamic models through bayesian forecasting are afordable with this book. Introductory chapters on the bayesian learning algorithm and univariate models rough out the kernel of the issue. Once you dive into the following more complicated chapters you can get lost but the main idea is got. To avoid getting lost, several readings are necessary. Finally, last chapters for non linear models, models with exponential distributions and MCMC methods are really heavy going but a light reading can allow you to get a general overview.

All in all, is a great workbook. The main drawback may be the lack of more practical examples to illustrate the theoretical concepts.

Each topic is thoroughly covered with theoretical rigor. I wish the authors publish an applied book with numerical example, and have all the algorithms coded in R or MATLAB. Their earlier attempt of black-box software BATS was entirely outdated, and black-box software sucks for its lack of flexibility.