Maximum Rank Correlation Estimation for Generalized Varying-Coefficient Models with Unknown Monotonic Link Function

Generalized varying coefficient models (GVCMs) form a family of statistical utilities that are applicable to real world questions for exploring associations between covariates and response variables. Researchers frequently fit GVCMs with particular link transformation functions. It is vital to recognize that to invest a model with a wrong link could provide extremely misleading knowledge. This thesis intends to bypass the actual form of the link function and explore a set of GVCMs whose link functions are monotonic. With the monotonicity being secured, this thesis endeavours to make use of the maximum rank correlation idea and proposes a maximum rank correlation estimation (MRCE) method for GVCMs. In addition to the introduction of MRCE, this thesis further extends the consideration to Generalized Semi-Varying Coefficient Models (GSVCMs), Panel data, simulations and empirical studies.