Clustering matrix-object data by correlational structure as proxy causal signals.

Matrix-object clustering addresses samples with multiple records per object. Many existing methods overlook within-object dependencies, which reduces interpretability and mixes heterogeneous regimes. We propose a clustering approach that uses intra-object correlational structure as a proxy for causal signals to separate regimes prior to any formal causal discovery. Each object is transformed into a rank-based correlation representation, enabling standard distance-based clustering while preserving interpretability. On synthetic and real-world datasets, the method yields stable, interpretable clusters that reduce regime mixing. We emphasize the boundary that correlation does not imply causation; correlational patterns are used only as proxy signals under the stated assumptions.