Hervé Abdi’s research while affiliated with The University of Texas at Dallas and other places

Similarity analyses between multiple correlation or covariance tables constitute the cornerstone of network neuroscience. Here, we introduce covSTATIS, a versatile, linear, unsupervised multi-table method designed to identify structured patterns in multi-table data, and allow for the simultaneous extraction and interpretation of both individual and group-level features. With covSTATIS, multiple similarity tables can now be easily integrated, without requiring a priori data simplification, complex black-box implementations, user-dependent specifications, or supervised frameworks. Applications of covSTATIS, a tutorial with Open Data and source code are provided. CovSTATIS offers a promising avenue for advancing the theoretical and analytic landscape of network neuroscience.

Reading storybooks aloud to prereaders helps them develop oral language skills; but educators struggle to select appropriate storybooks for read alouds. To better understand this selection process, we evaluated how experienced speech language pathologists (SLPs) apply the storybook selection rubric in Schwarz et al.'s (2015). A read-aloud storybook selection system for prereaders at the preschool language level: A pilot study.. That rubric includes: eight book characteristics, one four-level difficulty scale, and exemplar storybooks for each scale level. Using this rubric, 38 SLPs-who had served children at the preschool language level-rated 63 storybooks based on how difficult they thought the storybooks would be for children at the preschool language level to understand when the storybooks were read aloud to them. A principal component analysis identified-among the original eight characteristics a subset of five highly correlated characteristics related to overall text difficulty: vocabulary, story structure, sentence length, book length, and density. We revised the original rubric to include only four storybook characteristics: (a) text difficulty, (b) amount of inferencing, problem-solving, and abstract concepts, (c) familiarity of activities/experiences, and (d) level of support provided by the illustrations. Using quartile values of the text difficulty characteristic, we derived a four-level difficulty scale for the 63 storybooks. The revised storybook selection rubric simplifies the process of selecting appropriate storybooks for read alouds. The revised rubric and SLPs' difficulty ratings of 63 storybooks provide clinicians with a resource ready-made for clinical practice. ARTICLE HISTORY

Correspondence analysis, multiple correspondence analysis and their discriminant counterparts (i.e., discriminant simple correspondence analysis and discriminant multiple correspondence analysis) are methods of choice for analyzing multivariate categorical data. In these methods, variables are integrated into optimal components computed as linear combinations whose weights are obtained from a generalized singular value decomposition (GSVD) that integrates specific metric constraints on the rows and columns of the original data matrix. The weights of the linear combinations are, in turn, used to interpret the components, and this interpretation is facilitated when components are 1) pairwise orthogonal and 2) when the values of the weights are either large or small but not intermediate-a pattern called a simple or a sparse structure. To obtain such simple configurations, the optimization problem solved by the GSVD is extended to include new constraints that implement component orthogonality and sparse weights. Because multiple correspondence analysis represents qualitative variables by a set of binary variables, an additional group constraint is added to the optimization problem in order to sparsify the whole set representing one qualitative variable. This new algorithm-called group-sparse GSVD (gsGSVD)-integrates these constraints via an iterative projection scheme onto the intersection of subspaces where each subspace implements a specific constraint. In this paper, we expose this new algorithm and show how it can be adapted to the sparsification of simple and multiple correspondence analysis, and illustrate its applications with the analysis of four different data sets-each illustrating the sparsification of a particular CA-based analysis.

Canonical correlation analysis (CCA) and partial least squares correlation (PLS) detect linear associations between two data matrices by computing latent variables (LVs) having maximal correlation (CCA) or covariance (PLS). This study compared the similarity and generalizability of CCA- and PLS-derived brain-behavior relationships. Data were accessed from the baseline Adolescent Brain Cognitive Development (ABCD) dataset (N > 9,000, 9–11 years). The brain matrix consisted of cortical thickness estimates from the Desikan-Killiany atlas. Two phenotypic scales were examined separately as the behavioral matrix; the Child Behavioral Checklist (CBCL) subscale scores and NIH Toolbox performance scores. Resampling methods were used to assess significance and generalizability of LVs. LV1 for the CBCL brain relationships was found to be significant, yet not consistently stable or reproducible, across CCA and PLS models (singular value: CCA = .13, PLS = .39, p < .001). LV1 for the NIH brain relationships showed similar relationships between CCA and PLS and was found to be stable and reproducible (singular value: CCA = .21, PLS = .43, p < .001). The current study suggests that stability and reproducibility of brain-behavior relationships identified by CCA and PLS are influenced by the statistical characteristics of the phenotypic measure used when applied to a large population-based pediatric sample.

Science, technology, engineering and mathematical (STEM) occupations are crucial for economic growth and individual financial stability, yet there is a STEM labour shortage, particularly among women. We examined how individual differences in visual imagery relate to characteristics of STEM occupations — specifically those requiring computational abilities. In a discovery cohort of 2357 online participants, we found that — consistent with prior research — spatial thinking was positively associated with STEM occupations for both males and females. Object imagery (mnemonic vividness), however, was negatively associated with STEM occupations that require computational thinking, possibly because efficient analytical reasoning abilities associate with low object imagery. This negative association was present for males, but not for females. We extended these findings to a sample of 1891 individuals with aphantasia (congenitally low imagery) and a sample of 186 university undergraduates. These results suggest that the well-known influence of spatial imagery is evident across genders, whereas an independent influence of non-spatial and non-visual abstract analytic abilities on computational STEM professions is confined to males. These findings have implications for policy in fostering careers in STEM, particularly for females.

Processing in machine learning qualitative variables having a very large number of modalities is an opportunity to revisit the theory of optimal scaling and its applications. This revisitation starts with the pioneers of scaling in statistics, psychometrics and psychology before moving on to more contemporary treatments of scaling that fall within the realm of machine learning and neural networks.

Gestures are an important social communication skill that infants and toddlers use to convey their thoughts, ideas, and intentions. Research suggests that early gesture use has important downstream impacts on developmental processes, such as language learning. However, autistic children are more likely to have challenges in their gestural development. The current study expands upon previous literature on the differences in gesture use between young autistic and non‐autistic toddlers by collecting data using a parent‐report questionnaire called the MCDI–Words and Gestures at three time points, 12, 18, and 24 months of age. Results ( N = 467) showed that high‐likelihood infants who later met diagnostic criteria for ASD ( n = 73 HL‐ASD) have attenuated gesture growth from 12 to 24 months for both deictic gestures and symbolic gestures when compared to high‐likelihood infants who later did not meet criteria for ASD ( n = 249 HL‐Neg) and low‐likelihood infants who did not meet criteria for ASD ( n = 145 LL‐Neg). Other social communicative skills, like play behaviors and imitation, were also found to be impacted in young autistic children when compared to their non‐autistic peers. Understanding early differences in social communication growth before a formal autism diagnosis can provide important insights for early intervention.

Background: Prior work has shown that certain modifiable health, Alzheimer's disease (AD) biomarker, and demographic variables are associated with cognitive performance. However, less is known about the relative importance of these different domains of variables in predicting longitudinal change in cognition. Objective: Identify novel relationships between modifiable physical and health variables, AD biomarkers, and slope of cognitive change over two years in a cohort of older adults with mild cognitive impairment (MCI). Methods: Metrics of cardiometabolic risk, stress, inflammation, neurotrophic/growth factors, and AD pathology were assessed in 123 older adults with MCI at baseline from the Alzheimer's Disease Neuroimaging Initiative (mean age = 73.9; SD = 7.6; mean education = 16.0; SD = 3.0). Partial least squares regression (PLSR)-a multivariate method which creates components that best predict an outcome-was used to identify whether these physiological variables were important in predicting slope of change in episodic memory or executive function over two years. Results: At two-year follow-up, the two PLSR models predicted, respectively, 20.0% and 19.6% of the variance in change in episodic memory and executive function. Baseline levels of AD biomarkers were important in predicting change in both episodic memory and executive function. Baseline education and neurotrophic/growth factors were important in predicting change in episodic memory, whereas cardiometabolic variables such as blood pressure and cholesterol were important in predicting change in executive function. Conclusion: These data-driven analyses highlight the impact of AD biomarkers on cognitive change and further clarify potential domain specific relationships with predictors of cognitive change.