Conditionally linear, matrix normal state space models

Abstract

We develop a class of linear state space models for matrix-valued time series data where the state is a latent matrix normal process. We derive matrix versions of the Kalman filter, log-likelihood, and smoother enabling estimation of the latent state matrix as well as the model's parameters. To conduct Bayesian inference, we provide algorithms that draw from the joint posterior distribution of the latent state matrices conditional on the observed data and parameters. We apply these methods to a large panel of U.S. macroeconomic time series across the 50 U.S. states. The proposed framework accommodates mixed-frequency data, heteroskedasticity, and outliers within a unified matrix-valued structure. Empirically, we find that a small number of latent factors captures the joint dynamics across states and variables, providing a parsimonious and scalable approach to modeling high-dimensional macroeconomic systems.

High-order fractional linear differential equations do not describe stochastic processes with finite state space
P. Van Mieghem
arXiv2 Nov 2026PreprintCC BY-NC-ND 4.0Full textAnalysis & PDEs Dynamical systemsarXiv
A Design Space Study of Density Matrix Parameterizations for Diffusion-Based Quantum State Tomography
Shuangu Chang
arXiv19 Sep 2026PreprintCC BY 4.0Full textMathematical physics Quantum physics & quantum informationarXiv
PAC-Bayesian Bounds for Learning Partially Observed Stochastic Linear Time-Invariant State-Space Systems with Inputs and Sub-Gaussian Noise
Mihaly Petreczky, Mohamad Al Ahdab, John Leth
arXiv8 Sep 2026PreprintCC BY 4.0Full textOptimisation & controlarXiv
Efficient Convex Optimization Methods for State-Space Heartbeat Dynamics Models with Gamma Generalized Linear Models
Sabrina Liu, Andrew S. Perley, Todd P. Coleman
arXiv1 Oct 2026PreprintarXiv licenceFull textBioinformatics & computational biology StatisticsarXiv
Asymptotics for partition functions of random normal matrix models with singularities
Kohei Noda
arXiv21 Sep 2026PreprintCC BY 4.0Full textMathematical physics ProbabilityarXiv

Cite

BibTeX RIS

BibTeX

@misc{creal2026conditionally,
  title = {{Conditionally linear, matrix normal state space models}},
  author = {Drew D. Creal and Marcelo C. Medeiros and Rodrigo Sarlo},
  howpublished = {arXiv},
  year = {2026},
  month = sep,
  doi = {10.48550/arxiv.2609.18734},
  eprint = {2609.18734},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2609.18734},
  note = {CC BY 4.0},
}

RIS

TY  - UNPB
TI  - Conditionally linear, matrix normal state space models
AU  - Drew D. Creal
AU  - Marcelo C. Medeiros
AU  - Rodrigo Sarlo
PY  - 2026
DA  - 2026/09/16
PB  - arXiv
DO  - 10.48550/arxiv.2609.18734
UR  - https://arxiv.org/abs/2609.18734
AB  - We develop a class of linear state space models for matrix-valued time series data where the state is a latent matrix normal process. We derive matrix versions of the Kalman filter, log-likelihood, and smoother enabling estimation of the latent state matrix as well as the model's parameters. To conduct Bayesian inference, we provide algorithms that draw from the joint posterior distribution of the latent state matrices conditional on the observed data and parameters. We apply these methods to a large panel of U.S. macroeconomic time series across the 50 U.S. states. The proposed framework accommodates mixed-frequency data, heteroskedasticity, and outliers within a unified matrix-valued structure. Empirically, we find that a small number of latent factors captures the joint dynamics across states and variables, providing a parsimonious and scalable approach to modeling high-dimensional macroeconomic systems.
ER  -

Abstract

Related

Cite