Ahmad Farid Osman

Exponential smoothing with regressors: Estimation and initialization

The main objective of this paper is to outline the estimation and initialization procedures for the exponential smoothing with regressors forecasting approach which was recently introduced. The paper also discusses what restrictions need to be imposed during the estimation process so that the algorithm satisfies the forecastability conditions. An empirical study using real non-seasonal data shows that the new approach sometimes has the ability to produce better forecasts than the existing exponential smoothing methods without regressors.

Stability and Forecastability Characteristics of Exponential Smoothing with Regressors Methods

Quite recently, two separate methods of exponential smoothing augmented with regressors have been introduced. The inclusion of regressors in the exponential smoothing method is intended to provide explanatory effects to a dependent variable rather than only relying on historical movement of a single series in generating forecasts. The objective of this study is to investigate the stability and forecastability characteristics of the two methods. Satisfying either stability or forecastability conditions ensures the method to produce stable forecast values for long forecast horizons. The findings of the study suggest that one of the methods is able to satisfy the forecastability conditions while another method fails to satisfy both stability and forecastability conditions.

Parameter Estimation of the Exponential Smoothing with Independent Regressors Using R

As an extension to the state space approach of the exponential smoothing methods, Osman and King recently introduced a new class of exponential smoothing methods by integrating regressors into the model. For the model to be utilized successfully, it requires a proper estimation procedure. The parameter estimation can be done through optimization using the “optim” function available in R statistical software. The objective of this paper is to discuss the effective use of the “optim” function in estimating parameters of the state space model of the exponential smoothing method augmented with regressors. The study started by considering several sets of optimization R codes to be supplied to the “optim” function. These codes use different set of initial values as the starting points for the optimization routine. The other difference between optimization codes is also in terms of restrictions imposed on the optimization routine. The second phase of the study was done by generating a number of simulated time series data with predetermined parameter values. The final phase of the study was then conducted by applying the optimization codes on all simulated series. By analyzing the performance of each of the optimization codes to accurately estimate parameters, a guideline or suggestion on how to effectively execute “optim” function in R with respect to the use of the new forecasting approach then outlined.