Elementwise Precision-weighted Compositional Data Analysis by Weighted Alternating Least Squares
Jan Graffelman
Abstract
Logratio principal component is a multivariate method commonly used for exploring compositional data sets. This method is mostly used in an unweighted manner, employing the centred logratio transformation of the compositions. The smaller elements of a composition usually have a higher relative error, and can be downweighted using part-wise weighting. In this article we focus on weighting compositional data and propose elementwise precision weighting of compositions. This is particularly convenient for compositions with parts that fluctuate from rare to frequent; e.g. parts acting as traces in some samples, but being abundant in others. We suggest to weight in the logarithmic or logratio scale, with the inverse of the standard deviation of the logratio, the latter being estimated for each part of each composition separately. A weighted least squares algorithm is used for dimension reduction and compositional biplot construction. Several applications of elementwise precision weighting are described, using compositional datasets from genetics, archaeology and microbiomics that illustrate the power and advantages of a fully weighted compositional analysis.
Create a lesson
Related papers
Additive Fréchet Regression for Random Objects
Changwon Choi, Wookyeong Song, Hans-Georg Müller et al.
Realized versus Population Reliability: Anytime-Valid Self-Consistency Certification under Correlated Reasoning
Terence D. Agbeyegbe, Toneukarin G. Agbeyegbe
Spatiotemporal Risk-Based Hotspot Dynamics Assessment
José M. Angulo, Ana E. Madrid
Optimal Transport Rank Regression for Robust Multivariate-Response Regression
Yizhe Ding, Yifan Jiang, Runze Li et al.
Statistical Inference with Ill-Conditioned Random Weight Neural Network Sieves
Giovanni Ballarin, Lukas Gonon, Lyudmila Grigoryeva
Causal Effect Estimation Under Uncertain Statistical Surrogacy
Yi Liu, Alexander W. Levis, Shu Yang et al.