Multivariate Statistics

Geostatistical studies often consider many variables; geotechnical and metallurgical properties, porosity, permeability and size distributions. Multivariate transformations, such as principal component analysis, the stepwise conditional transform or projection pursuit multivariate transform may be applied for decorrelation or transformation to a multivariate Gaussian distribution. Multivariate probability distributions may be fit with kernel density estimates or Gaussian mixture models. Complex relationships may be understood with nonlinear regression techniques such as alternating conditional expectations.

Lessons

• Gaussian Mixture Models
• Cokriging with Unequally Sampled Data
• Aggregating Variables into a Super Secondary Variable
• Principal Component Analysis
• Multidimensional Scaling (see source code on GitHub)
• Projection Pursuit Multivariate Transform

• Combination of Multivariate Gaussian Distributions through Error Ellipses
• Multivariate Gaussian Distribution
• Sphereing and Min/Max Autocorrelation Factors