latent Gaussian copula model

The latent Gaussian copula model is a statistical tool for integrating and analyzing heterogeneous data types. It mainly focuses on the combination of functional magnetic resonance imaging (fMRI) and single nucleotide polymorphism (SNP) data for the comprehensive diagnosis of mental disorders. The model aims to uncover complex interactions among these diverse data types, providing a new perspective on the underlying neurogenetic mechanisms.

Key features of the latent Gaussian copula model:

1. Handles mixed data types: The model can integrate continuous, binary, count, and multinomial data, such as SNP data, which has been a challenge for existing graphical models.

2. Latent variable assumption: The model assumes that discrete variables are obtained by discretizing latent (unobserved) continuous variables, enabling the creation of a semi-rank-based estimator of the graph structure.

3. Accurate graph structure detection: Simulation results demonstrate that the proposed latent correlation has a more steady and accurate performance in detecting graph structures compared to several existing methods.

4. Biologically significant associations: When applied to real schizophrenia data consisting of an SNP array and fMRI images, the model reveals distinct SNP-brain associations verified to be biologically important.