Title: Multi-CHull for Multiverse Analysis: Studying Model Sensitivity by
Applying CHull with Multiple Fit Values

Authors: Jeffrey Durieux, Marre Vervloet, Eva Ceulemans, Tom F. Wilderjans

Abstract:
The CHull method, which was proposed by Ceulemans and Kiers (2006), is a
generic procedure for model selection that identifies the model that optimally
balances model fit and complexity. Several choices that typically have to be
made before and/or during the analysis (e.g., pre-processing strategy, choice
of model complexity definition) will have an influence on which model is
optimal. It is good practice for users to critically evaluate their choices
made in this regard and to investigate the effect of these choices on the
selected optimal model. To this end, users should run a Multi-CHull analysis
in which CHull is applied several times (e.g., determining the optimal model
for each pre-processing strategy). This approach of performing multiple
analyses, herewith using different settings each time such that effects or
outcomes of these choices can be evaluated, aligns with the concept of
multiverse analysis (Steegen, Tuerlinckx, Gelman, & Vanpaemel, 2016). To aid
the user in this, we translated the CHull software into an R package and
extended it with functions for a Multi-CHull analysis. In this presentation,
we, present the open source multichull R package which can be installed from
CRAN and an associated RShiny App
(https://multichull.shinyapps.io/multichull_shiny/) to assist the user in
applying the Multi-CHull procedure. The functionality of the package is
demonstrated with empirical data analysis applications.

Ceulemans, E., & Kiers, H. A. L. (2006). Selecting among three-mode principal
component models of different types and complexities: A numerical convex hull
based method. British Journal of Mathematical and Statistical Psychology, 59,
133-150.

Steegen, S., Tuerlinckx, F., Gelman, A., & Vanpaemel, W. (2016). Increasing
transparency through a multiverse analysis. Perspectives on Psychological
Science, 11, 702-712.