In this paper a Bayesian-testing procedure of the seasonality pattern (single and/or double) is proposed. The data generating process is assumed to be Double Seasonal Autoregressive (DSAR) model. The testing procedure is based on deriving the posterior probability distribution of the parameters then testing the significance of the corresponding seasonal parameters for seasonality pattern. For conjugate and Jeffreys priors, the testing procedure is based on the F distribution. The proposed Bayesian test is demonstrated by simulated and real life examples.
View on SCiNiTO
Top