The Analysis of Mixed Model in Time Series Decomposition

This paper examines the condition(s) under which the mixed model is the most appropriate model in descriptive time series analysis when trend-cycle component is linear. Since some existing studies have adequately characterized the additive model, the analyst is still exposed to the risk of wrongly using multiplicative model when mixed model should be used. Therefore, the aim of this study is to identify the series that admits mixed model. The method employed in this paper is the Buys-Ballot procedure developed for choice of model and choice of appropriate transformations, among other uses, based on row, column and over totals, means and variances of the Buys-Ballot table. Results show that, 1) the seasonal variance of the Buys-Ballot table, for the mixed model, a function of the slope and seasonal effect only. 2) the calculated values of the test statistic identified the mixed model correctly in 99 out the 100 stimulations