Stationary Unit Roots and Inference in Softplus INGARCH Models
math.ST, stat.TH
Submitted: 2026-10-14
Updated: 2026-11-02
License: http://creativecommons.org/licenses/by/4.0/
The gist: A unit root makes linear count time series models such as the Poisson INGARCH(1,1) nonstationary.
Terminology
Abstract
A unit root makes linear count time series models such as the Poisson INGARCH(1,1) nonstationary. We show that a softplus link changes this. Weiß, Zhu and Hoshiyar (2022) introduced softplus INGARCH models and gave a sufficient condition for stationarity. For Poisson, negative binomial, gamma and related versions of these models, we obtain the necessary and sufficient condition, which allows a unit root. At a stationary unit root, the conditional mean has a power-law tail, and the observations can have infinite variance or long memory. To our knowledge, this is the first stationary unit root established for an INGARCH-type count model. For a fixed softplus scale, the Poisson quasi-maximum likelihood estimator is asymptotically normal both below and at the unit root, including cases with infinite variance. Two applications show that unit-root conclusions can depend on the softplus scale.
Sources
- Near-Unit-Root Theory for Affine Processes
- Stability of discrete-time Hawkes process with inhibition: towards a general condition
Related papers
- Sigmoid-FTRL: Design-Based Adaptive Neyman Allocation for AIPW Estimators
- Nonparametric Regression on Fixed-Cardinality Subsets:Minimax Risk and Random-Design Effects
- Conformal Prediction for Dyadic Regression Under Complex Missingness
- Optimal Linear Functional Estimation for Proportionally High-Dimensional Linear Models
- Semiparametric Efficient Inference under Non-Informative Complex Survey Designs
- Sharp Asymptotic Theory of Maximum Likelihood Estimation for Gaussian Processes with an RBF Kernel