Date of Completion
Clustering, Model selection, LASSO
Field of Study
Doctor of Philosophy
In this study, we propose a cluster analysis method based on functional relationship between response variables and covariates. We assume the model with functional relationship specific to group c between variable y and covariate x as
y = G(g(x, beta_c), epsilon)
where random variable epsilon follows a group-specific distribution D(theta_c). We discuss several scenarios of the proposed model and error rates in the context of classification. Numerical examples are also given in this dissertation as well as the reference implementations in the appendix.
Liao, Gong-Yi, "Residual Likelihood Based Cluster Analysis" (2013). Doctoral Dissertations. 149.