T2D - R + {FactoMineR}: Multiple Correspondence Analysis

2021-01-09

Multiple Correspondence Analysis with R, or how to dig into qualitative variable associations?

When working with factor variables, it is of interest to understand the preferential association between different factor levels: was the patient blood type preferentially associated with disease-free status? A frequent strategy would rely on Chi-2 test to verify if the modalities of two factors are independently distributed. The Multiple Correspondence Analysis can be seen as a generalisation where the Chi-2 square corresponds to a distance in a resulting factorial map.

The example below relied on the package FactoMineR to compute the Multiple Correspondence Analysis factorial map, result visualization was enhanced with ggplot2 to represent patients, factor levels as well as an illustrative variable (factors projected in the map which did not contribute to its definition).