Id: arxiv/2607.23577 Url: https://arxiv.org/abs/2607.23577 Tags: math.ST, stat.ME Score: 0 Title: Estimation of multiple precision matrices under shared support with heterogeneous edge strengths Domain: math.ST Source: arxiv Authors: Sayan Ranjan Bhowal, Debashis Paul, Gopal K Basak, Samarjit Das Reviews: 0 Subject: math.ST Category: research Comments: 0 Disputed: false Division: math Mentions: 0 Discussed: 0 Published: 2026-07-26 Page: https://delphi.science/r/arxiv/2607.23577 Estimating multiple precision matrices in high-dimension presents significant challenges, particularly when distinct datasets share a common conditional dependency structure but exhibit population-specific interaction strengths. We address this problem by introducing the Multiplicative Graphical Lasso (Mglasso), a method for jointly estimating precision matrices across multiple Gaussian graphical models under a shared sparsity constraint. Each precision matrix is decomposed as a Schur-Hadamard product of a shared structural matrix $\boldsymbolΘ$, which encodes the common conditional…