Workshop on "Compositional data analysis”
by Dalel Abdi*, Serge-Étienne Parent§ & Léon-Etienne Parent§
* Agriculture and Agri-food Canada, Soils and Crops Research and Development Centre, Canada
§ Université Laval, Department of Soils and Agri-Food Engineering, Canada
Date: 5 July 2015 (1-5 PM)
Location: To be announced
Registration Fee: Free
Please contact Dalel Abdi at dalel.abdi@agr.gc.ca to register
Limited to 20 participants
Compositional data refer to proportions, or parts of some whole, bounded between 0 and the unit of measurement, i.e., 1, 100%, 1000 g kg−1, or 106 mg kg−1. Examples of compositions are data presented in ppm, ppb, molarities, or any other concentration units, such as soil, compost and crop analysis. The constrained nature of compositional data implies particular and important numerical properties that have major consequences for any statistical analysis. Standard techniques are designed to be used with data that are free to range from −∝ to +∝ (Pawlosky-Glahn and Egozcue, 2006). Any increase in one component must be associated with a decrease in one or more components in a closed system. This means that compositional data are intrinsically correlated to each other, and the results of standard statistical analysis such as regression, univariate and multivariate analysis are misleading (Aitchison, 1986). Hence, contradictories interpretations could be generated which raise a serious problem that needs to be treated with considerable circumspection. Compositional data analysis using additive log-ratio (alr), centred log-ratio (clr) or isometric log-ratio (ilr) coordinates avoids such difficulties and preserves sub-compositional coherence in the analysis (Aitchison, 1986; Egozcue et al, 2003). This workshop will provide examples of biased analysis obtained with agronomic data and present compositional approaches to deal with these problems using the “R” software. The session will be on Sunday 5th, July from 1 to 5 pm. Please contact Dalel Abdi at dalel.abdi@agr.gc.ca to register (the session can hold 20 persons).