Identifying Subtypes Among Children With Developmental Coordination Disorder and Mathematical Learning Disabilities, Using Model-Based Clustering

Abstract

A relationship between motor and mathematical skills has been shown by previous research. However, the question of whether subtypes can be differentiated within developmental coordination disorder (DCD) and/or mathematical learning disability (MLD) remains unresolved. In a sample of children with and without DCD and/or MLD, a data-driven model-based clustering was used to identify subgroups of individuals with relatively homogeneous profiles on measures associated with motor and mathematical skills. One subgroup of children with motor problems was found based on motor variables. Based on mathematical variables, two clinical clusters were found: a subtype with number fact retrieval problems and a subtype with procedural calculation problems. Clustering with motor and mathematical skills revealed two clinical clusters: a cluster with number fact retrieval as well as procedural calculation problems and below average motor and visual-motor integration skills. A second cluster of children had only procedural calculation and visual-motor problems. Our results raise questions about the usefulness of placing children who have below average mathematical skills into a single diagnostic category. Furthermore, we inform ongoing debates about the overlap between DCD and MLD, as below average motor skills were found in both MLD subgroups, although a different motor profile is linked to a different mathematical profile.

Mapping an Experiment‐Based Assessment of Collaborative Behavior Onto Collaborative Problem Solving in PISA 2015: A Cluster Analysis Approach for Collaborator Profiles

“…We identified the number of clusters (Research Question 1) by relying on relative fit indices (Fraley & Raftery, 2002;Geiser, 2011) as well as class attribution probabilities and interpretability with respect to theoretical considerations (Geiser, 2011). A cluster analysis assumes that the data integrated into the analysis were generated from a population that consisted of several subpopulations (Pieters, Roeyers, Rosseel, Van Waelvelde, & Desoete, 2015). A cluster analysis assumes that the data integrated into the analysis were generated from a population that consisted of several subpopulations (Pieters, Roeyers, Rosseel, Van Waelvelde, & Desoete, 2015).…”

“…With respect to relative fit indices, a model-based cluster analysis provides best model fit by calculating a number of Gaussian models per cluster, whereby a Bayesian information criterion (BIC) is determined for each model (Fraley & Raftery, 2002). A cluster analysis assumes that the data integrated into the analysis were generated from a population that consisted of several subpopulations (Pieters, Roeyers, Rosseel, Van Waelvelde, & Desoete, 2015). Each subpopulation (or cluster) is modeled by Gaussian models of different types of multivariate normal distributions.…”

Using the ICF Framework to Explore the Multiple Interacting Factors Associated with Developmental Coordination Disorder

“…Difficulty in applying knowledge is apparent in academic settings where children with DCD display difficulty in reading [3] , writing [33] and calculating [34]. Inability to focus attention and filter out distractions often compounds the problem in executing these skills as DCD often co-occurs with deficits in attention [3].…”

Cognitive heterogeneity of math difficulties: A bottom-up classification approach

“…The limitations of the top-down approach have led some researchers to pursue the data-driven or bottom-up approach, which consists of letting the groups emerge from multivariate techniques of classification, such as cluster analysis (Archibald, Cardy, Joanisse, & Ansari, 2013;Bartelet, Ansari, Vaessen, & Blomert, 2014;Gray & Reeve, 2016;Osmon, Smerz, Braun, & Plambeck, 2006;Peake, Jiménez, & Rodríguez, 2017;Pieters, Roeyers, Rosseel, Van Waelvelde, & Desoete, 2015;Reeve, Reynolds, Humberstone, & Butterworth, 2012;Vanbinst et al, 2015;von Aster, 2000). Ideally, subgroups of individuals with intragroup similarities and between-group differences on some criterion variables should emerge.…”