Statistical Analysis Protocol to Identify Homogeneous Subtypes in Clinical Data
- Data preparation (scale, adjust, feature selection, sample selection, exploratory data analysis)
- Mixture Model Clustering (based on R mclust)
- Model Selection (model ranking, BIC quantiles, top-N models)
- Characterization and Comparison of Models (visualize heatmaps, parallel coordinates, cross compare models)
- Validation of Subtypes (application specific inference methods: risk ratio, chi2, ...)
========= Data
- Clinical Outcomes (scores, ...)
- Magnetic Resonance Spectroscopy
========= Remarks:
- Rely on R MClust by A. Raftery & C. Fraley
- R SDisc is no longer maintained