2024-10-04

Ruiqi Liu, Ph.D., Department of Mathematics and Statistics, Texas Tech University

"Estimating the Number of Groups in Panel Data Models: A Penalty-free Perspective"

This paper considers linear panel data models with latent group structures when the number of groups is unknown. I propose a novel, penalty-free procedure to estimate the number of groups. Based on two newly designed measurements of within-group dissimilarity, two estimators are proposed by maximizing the dissimilarity ratio. Compared to existing methods, the estimators do not require any penalty parameters, making them easy to compute in practice. Their consistency is established, and the proof employs novel techniques beyond the standard uniform law of large numbers. Simulation studies using synthetic data and an application to a real-world dataset confirm the effectiveness of the proposed methods.

Stay connected TwitterFacebook LinkedIn YouTubeInstagram