Introduction
The core objective of the mvpaShiny is to identify latent structures and important variables in multivariate datasets, containing multiple supposedly explanatory variables (X) and response (y). mvpaShiny uses the functions embedded in the mvpa R package (multivariate pattern analysis) bundled in an R shiny web app. This shall facilitate dataset handling and the downstream data analysis by offering a simple-to-use graphical user interface. mvpaShiny offers various methods to unravel the datasets, namely using principal component analysis (PCA), confounder projection and projection to latent structures (alias partial least squares, PLS) regression.
Objectives tackled by mvpaShiny
- Dataset preparation and inspection → Creation of subsets, scaling/transformation and initial inspection (correlation / variable normality)
- Reduction of dimensions / Handle collinearity → Substitute redundant variables by a new latent variable that combines multiple similar variables
- Adjust for covariate variables and alleviate their influence
- Create validated partial least squares (PLS) regression models and identify important variables → Build a formula of the given explanatory variables (X) that predicts and explains the response (y) as good as possible
- Visualization of association patterns to enhance model interpretation
- Integrate the above mentioned methods in a simple manner to handle complex datasets
Requirements
mvpaShiny can be used online using the following link mvpaShiny - web version
However, mvpaShiny can also be operated locally through R Studio.
An installation guide for the local version is offered here: How to install mvpaShiny