Abstract
This paper looks at a novel type of seasonality labeled as a seasonal Generalized auto-regressive (GAR) model. The seasonal GAR models are found to be short-memory models, and expressions for the autocorrelation function and large sample results for the parameter estimates are established. Traditional Box-Jenkins seasonality models and Gegenbauer seasonality models are compared with the seasonal GAR model. Finally, the three methods are compared in the analysis of a specific process - the Mauna Loa CO2 data - showing that in this case, the seasonal GAR model provides forecasts with a lower mean squared error.
KEYWORDS:
Mathematics Subject Classification::
Disclosure statement
No potential conflict of interest was reported by the authors.
Availability of data and material
Where a source of data is not specifically indicated in the paper, all data may be obtained from the authors.
Code availability
All code may be obtained from the corresponding author.