Stabilisation of the R package LMWiRe – Linear Models for Wide Responses

In a data-driven society, the development of efficient data analysis methods has become crucial. For example, in life sciences, and particularly in the “omics” discipline, various analytical chemistry techniques are frequently used to collect data. These techniques often produce spectra, which are high dimensional multivariate data where the number of response variables tends to be much larger than the number of samples analysed. This information is collected in a large but short matrix. Moreover, these experiments are often underpinned by multifactorial experimental designs. It is therefore in this context that the methodologies of ASCA+ (ANOVA-simultaneous component analysis) and APCA+ (ANOVA-principal component analysis) were developed. These methods are a combination of a parallel statistical modelling that allows each response of the outcomes matrix to be analysed separately, and a dimension reduction. The aim of this master’s thesis is to complete and stabilise the R package LMWiRe (Linear Models for Wide Responses) which allows the statistical modelling of wide response matrices using the ASCA+ method. This document serves, therefore, two purposes. On the one hand, it intends to demonstrate the work accomplished during this thesis and on the other hand, it aims to provide a user manual as well as the necessary tools for further development of this project. Using two case studies, this thesis illustrates how the package works by detailing the functions used at each step of the analysis and by statistically interpreting the outputs generated.