lmdme : Linear Models on Designed Multivariate Experiments in R

• Published in: Journal of statistical software Vol. 56; no. 7; pp. 1 - 16
• Main Authors: Fresno, Cristóbal, Balzarini, Mónica G., Fernández, Elmer A.
• Format: Journal Article
• Language: English
• Published: Foundation for Open Access Statistics 2014
• ISSN: 1548-7660; 1548-7660
• DOI: 10.18637/jss.v056.i07

The lmdme package decomposes analysis of variance (ANOVA) through linear mod- els on designed multivariate experiments, allowing ANOVA-principal component analysis (APCA) and ANOVA-simultaneous component analysis (ASCA) in R. It also extends both methods with the application of partial least squares (PLS) through the specification of a desired output matrix. The package is freely available from Bioconductor and licensed under the GNU General Public License. ANOVA decomposition methods for designed multivariate experiments are becoming popular in omics experiments (transcriptomics, metabolomics, etc.), where measurements are performed according to a predefined experimental design, with several experimental factors or including subject-specific clinical covariates, such as those present in current clinical genomic studies. ANOVA-PCA and ASCA are well-suited methods for studying interaction patterns on multidimensional datasets. However, currently an R implementation of APCA is only available for Spectra data in the ChemoSpec package, whereas ASCA is based on average calculations on the indices of up to three design matrices. Thus, no statistical inference on estimated effects is provided. Moreover, ASCA is not available in an R package. Here, we present an R implementation for ANOVA decomposition with PCA/PLS analysis that allows the user to specify (through a flexible formula interface), almost any linear model with the associated inference on the estimated effects, as well as to display functions to explore results both of PCA and PLS. We describe the model, its implementation and two high-throughput microarray examples: one applied to interaction pattern analysis and the other to quality assessment.