Abstract
This paper introduces a new stationary first-order autoregressive integer-valued process with a marginal distribution following the New Discrete Lindley distribution, referred to as NDLINAR( 1). The process is developed to model over-dispersed count time series exhibiting a mixed behavior arising from the combination of multiple distributions. Its statistical properties are thoroughly investigated, and the parameters are estimated using conditional maximum likelihood. The asymptotic properties of the estimators are also analyzed. The performance of the proposed model is evaluated by comparing it with other existing INAR(1) processes through applications to real datasets. Results demonstrate that the NDL-INAR(1) process effectively captures the characteristics of over-dispersed time series and provides improved predictive accuracy in comparison with alternative models. The study highlights the flexibility and applicability of the NDL-INAR(1) process for practical time series modeling of count data in various fields.
Recommended Citation
Kabore, Tégawendé Martin and Ouédraogo, Jean-Etienne Ouindllassida
(2050).
(R2174) A Stationary First-Order Autoregressive Process With New Discrete Lindley Marginal Distribution,
Applications and Applied Mathematics: An International Journal (AAM), Vol. 21,
Iss.
1, Article 6.
Available at:
https://digitalcommons.pvamu.edu/aam/vol21/iss1/6