![IconGems-Print](images/IconGems-Print.png)
581 – Advancement in Theoretical and Applied Aspects of Modeling
Revisiting the Linear Models With Exchangeably Distributed Errors
Timothy Opheim
The University of Texas at San Antonio
Anuradha Roy
The University of Texas at San Antonio
The popularity of the classical general linear model (CGLM) is mostly due to the ease of model building and authentication. However, the CGLM is not appropriate and thus not applicable for correlated two dimensional observations. In this article we revisit Arnold's (1979) exchangeable general linear model (EGLM) for one observation, derive profiled maximum likelihood estimates (P-MLEs) of the model's parameters, and obtain its joint complete sufficient statistics. We also obtain the joint complete sufficient statistics of the "extended" model of Arnold (1979) for the case of multiple observations.