Efficiently handling high‐dimensional data from multifactorial designs with unequal group sizes using Rebalanced ASCA (RASCA)
Open Access
- 2 May 2022
- journal article
- research article
- Published by Wiley in Journal of Chemometrics
- Vol. 37 (7)
Abstract
A novel chemometric approach is proposed to analyze high‐dimensional data from unbalanced designs of experiments. It combines a rebalancing strategy based on averages with the ASCA method under the name Rebalanced ASCA (RASCA). The ability of RASCA to handle unbalanced designs was compared with standard ASCA, as well as state‐of‐the‐art methods, such as ASCA+ and WE‐ASCA. For that purpose, a controlled framework was designed to provide a systematic comparison of the various approaches. It included two real datasets obtained from initially balanced designs, which were gradually unbalanced by removing observations belonging to specific combinations of factor levels. The results illustrate that all methods considered led to identical solutions when the initial balanced design was kept. Nevertheless, increasing differences appeared when the design was gradually unbalanced. The proposed benchmark showed that RASCA and ASCA+ provided overall similar results for all effects with high agreement with the balanced solutions in comparison to classical ASCA and WE‐ASCA. RASCA was found to be a suitable chemometric tool to tackle unbalanced designs by ensuring unbiased parameter estimators with the added benefit of producing rigorously orthogonal effect matrices, thus facilitating interpretation.Keywords
This publication has 28 references indexed in Scilit:
- Improving the analysis of designed studies by combining statistical modelling with study design informationBMC Bioinformatics, 2009
- Matrix correlations for high-dimensional data: the modified RV-coefficientBioinformatics, 2008
- Metabolite-based clustering and visualization of mass spectrometry data using one-dimensional self-organizing mapsAlgorithms for Molecular Biology, 2008
- PARAFASCA: ASCA combined with PARAFAC for the analysis of metabolic fingerprinting dataJournal of Chemometrics, 2008
- Statistical validation of megavariate effects in ASCABMC Bioinformatics, 2007
- Simultaneous clustering of gene expression data with clinical chemistry and pathological evaluations reveals phenotypic prototypesBMC Systems Biology, 2007
- ASCA: analysis of multivariate data obtained from an experimental designJournal of Chemometrics, 2005
- Analysis of variance–principal component analysis: A soft tool for proteomic discoveryAnalytica Chimica Acta, 2005
- ANOVA-simultaneous component analysis (ASCA): a new tool for analyzing designed metabolomics dataBioinformatics, 2005
- Permutation tests for multi-factorial analysis of varianceJournal of Statistical Computation and Simulation, 2003