Skip to Main content Skip to Navigation
Journal articles

Overcoming the Spurious Groups Problem in Between-Group PCA

Abstract : Several papers have recently raised the occurrence of some problems with between-group Principal Component Analysis (bgPCA). This method inflates the differences between the groups, and can even display completely artificial differences when none exist, for example when applied to random numbers tables with many variables (columns) and few individuals (rows). Lately, cross-validation has been proposed as a way to circumvent this problem. Here we present some tools and several functions of the ade4 package for the R statistical software to compute a bgPCA, test the presence of statistically significant groups, perform a cross-validation of this analysis and compute associated statistics. We also describe how to use these functions to avoid running into the spurious groups problem. Several examples, including a real data set and random numbers tables, are used to validate this approach in var
Complete list of metadata

https://hal.archives-ouvertes.fr/hal-03348141
Contributor : Jean Thioulouse Connect in order to contact the contributor
Submitted on : Friday, September 17, 2021 - 6:39:56 PM
Last modification on : Tuesday, October 19, 2021 - 9:48:03 AM

File

ThioulouseEB2021.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-03348141, version 1

Collections

Citation

Jean Thioulouse, Sabrina Renaud, Anne-Béatrice Dufour, Stéphane Dray. Overcoming the Spurious Groups Problem in Between-Group PCA. Evolutionary Biology, Springer, 2021. ⟨hal-03348141⟩

Share

Metrics

Record views

72

Files downloads

201