Compositional Data Analysis (CODA)

Abstract

Compositional data analysis is a methodology used to describe the parts/compounds of a whole, conveying relative information. Typical examples in different fields are: geology (geochemical elements), medicine (body composition: fat, bone, lean), food industry (food composition: fat, sugar, etc), chemistry (chemical composition), ecology (abundance of different species), agriculture (nutrient balan

Objective

Students will be able to: - decide where (and where not) methods for analyzing compositional data can be used - describe what their properties are and what challenges are associated with them, and to decide which method to choose for their research task - critically evaluate the model results of a compositional data approach in the context of plant science.

Content

The objective of this course is to introduce students with a basic programming background to compositional data analysis. We will discuss topics like the geometric properties of compositional data in plant science including the representation of data in so-called log-ratio coordinates, explanatory data analysis and visualization, location and covariance measures, application to multivariate analysis (e.g. cluster analysis), linear models and we give an outline on problems for high-dimensional data. In addition, problems with missing values, zeros and outliers are discussed. The course will consist of 50% lectures and 30% hands-on programming in R, where students will directly apply methods in software to help solving problems in plant sciences, and 20% is spent on a given task.

Resources