# Data Transformation

Many geostatistical techniques are constructed in the context of multivariate Gaussian data. Rarely are data multivariate Gaussian; data transformation is required. The widely used univariate quantile transform or probit enforces univariate Gaussianity. Linear rotations, such as principal component analysis and minimum/maximum autocorrelation factors, may be applied to decorrelate variables. Transformations such as the stepwise conditional transform and projection pursuit multivariate transform may be used to enforce multivariate Gaussianity. Compositional transforms, such as log-ratios, may be applied to remove sum constraints. These data transformations enable the application of geostatistical techniques which could not otherwise be applied. These lessons focus on the integration of data transformation techniques into geostatistical workflows and the decisions and assumptions which accompany their application.

# Lessons

- Hierarchical Transformation Workflows
- Multivariate Linear Rotation Transformations
- Multivariate Normal Transformations

- Cell Declustering Parameter Selection
*On Scale, Support, Resolution, Upscaling and Downscaling*