Advanced Time Series Data Analysis

1Forecasting a Monthly Time Series

1.1 Introduction

It is recognized that all possible models of a single time series, Y_t, can easily be applied to a forecast, Y_t. Refer to Agung (2009a), which presents a number of time series models that could easily be extended to many more possible models, as well as models based on panel data presented in Agung (2014), which can be used in forecasting. So, a researcher should never have to present the best possible forecasting, which should be highly dependent on his/her subjective expert judgment.

This chapter specifically presents forecasting based on a single monthly time series, namely Y_t, without taking into account the effects of exogenous variables, except for any lags or the time variable. More alternative and advanced models will be presented in the following chapters. For illustration, this chapter only presents selected illustrative forecasting based on the data in House.wf1, which contains only one single time series variable, namely HS_t, with 604 time‐observations from 1946M01 to 1999M04. In addition, for comparison, illustrative examples are presented based on other selected data sets.

1.2 Forecasting Using LV(p) Models

1.2.1 Basic or Regular LV(p) Models

It is well known that the LV(p) model of a time series variable, Y_t, has the following general form.

(1.1)

therefore, a forecast of any transformed variable G(Y_t) can easily ...