Efficient accounting for estimation uncertainty in coherent forecasting of count processes.

Area of Science:

• Statistics
• Time Series Analysis
• Computational Statistics

Background:

• Coherent forecasting for count processes generates discrete count forecasts.
• Forecasts are impacted by estimation uncertainty from fitted models.
• This uncertainty means forecasts may deviate from true values.

Purpose of the Study:

• To develop a computationally efficient resampling scheme for expressing uncertainty in coherent count process forecasts.
• To assess the performance of this scheme via simulation.
• To demonstrate its practical application with real data.

Main Methods:

• A novel resampling scheme is proposed.
• The scheme generates ensembles of forecast values.
• Performance is evaluated through simulation studies and a real-data example.

Main Results:

• The resampling scheme effectively expresses uncertainty in coherent count process forecasts.
• Simulation studies confirm the scheme's performance.
• The method allows for intuitive visual presentation of forecast ensembles.

Conclusions:

• The proposed resampling scheme provides a practical and interpretable way to handle forecast uncertainty in count processes.
• This approach enhances the reliability and understanding of count data forecasting.
• Visualizations of forecast ensembles aid in practical decision-making.