Power transformations of relative count data as a shrinkage problem

DOI10.1007/S41884-023-00104-1MaRDI QIDQ6170233zbMATH OpenOpenAlexFDO

Authors Ionas Erb

Publication date 12 July 2023

Published in Information Geometry (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/2205.09215

zbMATH Keywords

Box-Cox transformationcompositional datadual geodesicsempirical Bayesinformation geometryJames-Stein shrinkagemultinomial distributionzero handling

Mathematics Subject Classification ID

Multivariate analysis (62Hxx) Linear inference, regression (62Jxx) Communication, information (94Axx)

Abstract: Here we show an application of our recently proposed information-geometric approach to compositional data analysis (CoDA). This application regards relative count data, which are, e.g., obtained from sequencing experiments. First we review in some detail a variety of necessary concepts ranging from basic count distributions and their information-geometric description over the link between Bayesian statistics and shrinkage to the use of power transformations in CoDA. We then show that powering, i.e., the equivalent to scalar multiplication on the simplex, can be understood as a shrinkage problem on the tangent space of the simplex. In information-geometric terms, traditional shrinkage corresponds to an optimization along a mixture (or m-) geodesic, while powering (or, as we call it, exponential shrinkage) can be optimized along an exponential (or e-) geodesic. While the m-geodesic corresponds to the posterior mean of the multinomial counts using a conjugate prior, the e-geodesic corresponds to an alternative parametrization of the posterior where prior and data contributions are weighted by geometric rather than arithmetic means. To optimize the exponential shrinkage parameter, we use mean-squared error as a cost function on the tangent space. This is just the expected squared Aitchison distance from the true parameter. We derive an analytic solution for its minimum based on the delta method and test it via simulations. We also discuss exponential shrinkage as an alternative to zero imputation for dimension reduction and data normalization.

Cites work

• A geometric view of conjugate priors
• A logistic normal multinomial regression model for microbiome compositional data analysis
• A statistical perspective on the challenges in molecular microbial biology
• Bayesian inference for categorical data analysis
• Compositional data: the sample space and its structure
• Conjugate priors for exponential families
• Elements of Information Theory
• Entropy inference and the James-Stein estimator, with application to nonlinear gene association networks
• Estimation with quadratic loss.
• Geometry of the probability simplex and its connection to the maximum entropy method
• scientific article; zbMATH DE number 3122730 (Why is no real title available?)
• scientific article; zbMATH DE number 4128217 (Why is no real title available?)
• scientific article; zbMATH DE number 3251902 (Why is no real title available?)
• Information geometry and its applications
• On the Admissible Estimators for Certain Fixed Sample Binomial Problems
• Power transformations in correspondence analysis
• Statistical decision theory and Bayesian analysis. 2nd ed
• Statistical Interpretation of Species Composition
• Stein's Estimation Rule and Its Competitors--An Empirical Bayes Approach
• Theory of probability. A critical introductory treatment. Translated from the Italian by Antonio Machí and Adrian Smith. With a new preface by Adrian Smith. In 2 volumes

Cited in

(1)

• The chiPower transformation: a valid alternative to logratio transformations in compositional data analysis

This page was built for publication: Power transformations of relative count data as a shrinkage problem

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6170233)