Covariate Adaptive Clustering
Make Cluster Assignments
Creating k-fold Cross-Validation Groups
Create Principal Component Analysis (PCA) scores matrix
Create matrix of Thin-Plate Regression Splines (TPRS)
Multinomial Logistic Regression
Measures of Prediction Performance
Prediction of Cluster Membership
Covariate Adaptive Clustering
Predictive K-means Clustering
Cross-validation of Predictive K-means Clustering
Re-order cluster labels
Implements the predictive k-means method for clustering observations, using a mixture of experts model to allow covariates to influence cluster centers. Motivated by air pollution epidemiology settings, where cluster membership needs to be predicted across space. Includes functions for predicting cluster membership using spatial splines and principal component analysis (PCA) scores using either multinomial logistic regression or support vector machines (SVMs). For method details see Keller et al. (2017) <doi:10.1214/16-AOAS992>.