Beyond Local Independence: High-Dimensional Latent Class Graphical Models with Shared Block Structure

Seunghyun Lee, Yuqi Gu · 2026

Read Beyond Local Independence: High-Dimensional Latent Class Graphical Models with Shared Block Structure by Seunghyun Lee, Yuqi Gu (2026) 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-2606-29631.txt.

Description

Lee and Gu extend latent class models, tools that sort survey or genetic responses into hidden groups. The classical model assumes local independence: within a class, items do not influence each other directly. That premise is often unrealistic, since responses within a group clearly interact.

Their latent class graphical model keeps the clarity of the classical approach through a shared block grouping of the items. Within each block, class-specific dependence is allowed. A three-step estimator recovers classes by spectral clustering, then estimates covariance matrices and blocks, then fits sparse precision matrices. Error bounds cover every stage. Tests recover the structure well. Uses on election survey and HapMap3 genotype data reveal local dependence.

Catalog record

CREATORSeunghyun Lee, Yuqi Gu
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE119 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.29631

Browse the full library · About the Tower of Babel