Intercept parameter variance in count models
Background: Integer-valued time series models, such as Poisson autoregressive (PAR) and Self-Excited Threshold Poisson autoregressive (SETPAR) models, often exhibit high variance in their parameter estimates, particularly for intercept parameters.
Question / Future Work: Investigate the theoretical cause of the observed large variance-to-mean ratio in intercept parameter (omega) estimates across various count time series models, including the PAR, SETPAR, and classical GARCH(1,1) models, which currently lacks a formal explanation in the literature.
Why It Matters: Understanding this persistent estimation issue is crucial for improving the robustness and interpretability of count time series models, which are widely used in financial and public health forecasting.
Metadata & Links
- created_at
- 2026-04-16T05:46:15Z