DUGi

Compositional Data Analysis with Red-R

http://dugi.udg.edu/item/http:@@@@hdl.handle.net@@2072@@273632

The compositional analyst must use a series of software to transform raw compositionaldata and run statistical analyses on them. Tools for compositional data analysis areavailable in R, an open source widely-used statistical computing environment. However,using R requires prior programming knowledge. Red-R is an open-source, user-friendlyvisual data flow interface based on R. The interface uses principles of pipelineprogramming where functions are represented as icons, termed widgets, and data flowsfrom one function to another by drawing lines between them on a canvas. Red-R is ableto perform common data analysis tasks (hypothesis tests, analysis of variance,regressions, principal component analysis, data cloud plots, bar plots, biplots, etc.). Wehave developed a novel Red-R package which implements the compositions package inR. Our compositions package can be used to perform compositional data operations overraw data (closure, additive, centered and isometric log ratio transformations,perturbations and powering, etc.), and create compositional plots (ternary diagrams, ilrdendrograms,etc.) without prior programming knowledge, after few basic operations.The objective of this work is to present Red-R and its compositions package using anapplication example for geochemical data. The network of widgets provides an easy-tofollowstep-by-step procedure to run a large number of operations available in R, hencefacilitating the tasks of the compositional data analyst. Furthermore, the entire analysisnetwork can be saved and reloaded. Reports can be generated from the widget network todocument and share results. Non-programmers can have an easy access to the advancedtools available in compositions analysis

Universitat de Girona. Departament d’Informàtica i Matemàtica Aplicada