software

---

FactoMineR

FactoMineR, developed with Husson, F. and Lê, S., is an R package designed for principal component methods, offering tools to explore, summarize, and visualize data. Its dimensionality reduction techniques include:

• PCA: Principal Component Analysis for continuous data.
• CA: Correspondence Analysis for count data (e.g., document-word matrices).
• MCA: Multiple Correspondence Analysis for categorical data.
• FAMD: Factorial Analysis of Mixed Data for datasets combining continuous and categorical variables.

Additionally, FactoMineR includes specialized methods for analyzing groups of variables (e.g., Multiple Factor Analysis, MFA), groups of individuals, and hierarchical structures.
[FactoMineR website] – [CRAN package] – [Youtube video] – [JSS paper] – [To ask questions ] – [MOOC] – [Lecture on MFA]

MissMDA

developed with Husson, F., is an R package designed to handle missing values in multivariate data analysis. It enables the application of PCA, MCA, FAMD, and MFA directly on incomplete datasets. The package also provides tools for single and multiple imputation of continuous, categorical, and mixed data, using principal component methods to ensure accurate imputations. Blog posts: Multiple imputations;  Can we believe in the  imputation?
[missMDA website] – [CRAN package] – [Youtube video] – [JSS paper]

denoiseR

denoiseR is an R package designed to estimate low-rank matrices from noisy data. It implements various regularized versions of Singular Value Decomposition (SVD), including soft-thresholding of singular values, nonlinear transformations, and adaptive shrinkage techniques. The package also features the Iterative Stable Autoencoder, a novel approach for denoising count data.
[CRAN package] – [denoiseR paper]