Article ID: | iaor19981725 |
Country: | Portugal |
Volume: | 17 |
Issue: | 2 |
Start Page Number: | 201 |
End Page Number: | 217 |
Publication Date: | Dec 1997 |
Journal: | Investigao Operacional |
Authors: | Nunes M. Helena Mourio Silva |
Keywords: | time series & forecasting methods |
The general behaviour of the gasoline sales contains calendar effects due to changing month length, weekly periodicity's, and holidays (which represents the calendar effects). Moreover, it is important to model the impact of extended weekends, ‘Carnival’ and ‘Easter’ effects. Finally, it became necessary to seasonally adjust the remaining series. The number of Saturdays, Sundays and holidays does not change the volume of sales, but the working-days already have an impact. The additional analysis of the precipitation did not reveal a significant effect on sales. After the adjustment, a much smoother series is obtained, especially in terms of homologous growth rates.