Associate Professor Xinyan Fan's Team Publishes in JASA on Network Varying Coefficient Models
A research team led by Associate Professor Xinyan Fan from the School of Statistics at Renmin University of China has published a paper in the prestigious Journal of the American Statistical Association (JASA). The paper introduces a novel network-varying coefficient model, which innovatively extends traditional varying coefficient models to accommodate network-structured data.
The model treats regression coefficients as functions of latent “locations” on a network and employs a projected gradient descent algorithm to estimate both network parameters and coefficient matrices. It uses Bayesian information criteria for model selection and includes a penalized method to identify variables with significant varying effects.
The approach is validated through simulations and a real-world application in finance, showing improved flexibility and accuracy in handling complex dependency structures in data.