A Matrix Approach to Likelihood Evaluation and Signal Extraction for ARIMA Component Time Series Models

Written by:
RR88-22

Introduction

Three common approaches to evaluating (Gaussian) likelihoods and doing other computations with time series models might be called the classical approach, the Kalman filter approach, and the matrix approach. The classical approach works directly with difference equation forms of models (particularly for autoregressive - integrated - moving average (ARIMA) models) and such things as covariance generating functions and spectral densities.

Page Last Revised - October 28, 2021