Matrix-variate integer-valued autoregressive processes

Nuo Xu · 2025

Read Matrix-variate integer-valued autoregressive processes by Nuo Xu (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-07347.txt.

Description

Nuo Xu and 2 co-authors propose the first model in 2025 for time series arranged as matrices of whole numbers. Such data arise in sociology and economics, for instance counts laid out by region and category over time. No earlier work had modeled matrix-valued count series directly, leaving analysts to flatten or approximate them.

The new autoregressive model, where each period depends on earlier ones, uses two thinning operators that act from the left and right of the matrix to keep values whole. The authors study the model's properties and give two estimation methods, projection and iterative least squares, with their large-sample guarantees. A rule for choosing the model's order is also supplied. In simulations and comparisons, the model beats both continuous matrix models and flattened count models on this data type.

Catalog record

CREATORNuo Xu
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE61 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-07347

Browse the full library · About the Tower of Babel