PropertyValue
is nif:broaderContext of
nif:broaderContext
is schema:hasPart of
schema:isPartOf
nif:isString
  • Forty-five participants (70–164 months; Mage = 110.89; 34 male) with primary diagnoses of autism spectrum disorders (ASD; 14.9%), Attention Deficit/Hyperactivity Disorder (ADHD; 34%), Learning Disorder/Intellectual Disorders (LD/ID; 40.4%), or Anxiety (2.1%) were drawn from a neuropsychological clinic-referred sample. They constitute the complete subset of a larger clinic-referred sample (see below) that completed the measures relevant to the present investigation. Participants’ data used for this study was identified from among a sample of evaluations of children who were referred to the Pediatric Neuropsychology Service at the University of Chicago between 2002–2013 (N = 1500). Referrals for evaluation were typically received from pediatricians and treating specialty physicians to obtain a comprehensive assessment of the child’s neuropsychological status within the context of clinical concerns regarding functioning and learning. Children between the ages of two and emerging adulthood, with challenges affecting learning, behavioral regulation, emotional functioning, or social skills, comprise the broader clinic sample. All evaluations were conducted in the outpatient service and were directed and supervised by a licensed clinical psychologist with specialty training in pediatric neuropsychology. Tests selected for the assessment were based on clinical referral questions, and were administered and then scored by either a trained MA-level full-time psychometrician, or a graduate student trainee in clinical psychology who was certified as a psychometrician for the Service. Rechecking of scoring was undertaken for all cases. When warranted, formal clinical diagnoses were established utilizing DSM-IV-TR criteria. Families and referral sources were provided with a final report to guide intervention of academic, emotional/behavioral, and social needs. At the first appointment, the child’s parent or legal guardian was presented with a written consent for their child’s participation in the clinical evaluation, and for their allowance of deidentified data from the assessment to be utilized in ongoing service research. Children utilized for this study were all from among those whose parents consented to both the assessment and data research use; consenting process and study procedures were approved by the University of Chicago Institutional Review Board. As noted above, given the nature of the referral, all children included in this study were presented with a clinical neuropsychological battery that included assessment of their intellectual functioning (see below), language expression and comprehension, verbal and visual memory, visuospatial processing and sensorimotor abilities, attention, executive functioning, and emotional and behavioral status. Relevant to the study, all children received the measures discussed below; this led to the reduction in available participants for the study, given tests administered and diagnoses of concern. That is, the sample included here represents 100% of those who completed all of the measures described below. Participants completed the Developmental Neuropsychological Assessment, Second Edition (NEPSY-II) [68] subscales focusing on affect decoding (Affect Recognition), social memory (Memory for Faces), motor skills (Fingertip Tapping), visuomotor skills (Imitating Hand Positions), response inhibition (Arrows), attention & set-shifting (Auditory Response Set), verbal comprehension (Comprehension of Instructions). This measure has been well-validated (including during the norming stage) in the diagnostic groups employed in the present study, and employs standard scores that adjust for norms by age [51, 68–70]. Participants also completed standard IQ measures from either the Wechsler [71–73] or Differential Ability Scales [74] batteries. Mothers completed the Behavior Assessment System for Children, Second Edition (BASC-2) [75], which is well-normed across the age ranges and diagnostic groups used here. The BASC-2 Atypicality and Social Skills scales are key indicators of populations with common social deficits (e.g., ADHD and ASD) [76, 77], were analyzed. We first conducted preliminary analyses to detect deviations from normality, and judged whether extreme skewness and/or kurtosis precluded executing parametric analyses. We then computed bivariate correlations between measures to assess presence of sufficient collinearity to support the use of (PLSR). That is, PLSR requires all or most independent variables to be substantially intercorrelated in order to effectively obtain reliable estimates of combinations among them [64]. This is similar to factor analysis in that bivariate correlations must be obtained prior to multivariate estimates. Given the potentially incommensurate nature of the scales by which the predictors and outcomes are calculated, both parametric (Pearson’s r) and nonparametric (Spearman’s rho) correlations were conducted. Next, as is frequently done when deriving PLSR models [78], all variables were converted to Z-scores to facilitate appropriate scaling of derived linear combinations in PLSR. To test our first exploratory hypothesis, that linear combinations of NEPSY-II variables related to affective processing (Affect Recognition, Memory for Faces), motor planning and execution (Fingertip Tapping, Imitating Hand Positions), inhibition (Arrows, Auditory Response Set), and linguistic processing (Comprehension of Instructions) could be used to optimize cross-sectional prediction of BASC-2 mother-reported child Atypicality, we fit a corresponding PLSR model. We then evaluated the appropriate number of independent variables by examining the Root Mean Squared Error of Prediction (RMSEP) of each PLSR component, using Random Segment Cross-Validation on the entire sample, seeking the smallest value after the intercept. Next, we examined the loadings for each of the NEPSY-II variables to theoretically define each obtained component. Next, we examined the amount of variance in the cross-sectional predictor variables explained by the chosen component to ensure it accounted for a large proportion (qualification of variance estimates completed in accordance with Cohen’s recommendations) [79]. Next, we specified the corresponding Ordinary Least Squares (OLS) regression model with the specified number of components, as well as a Backwards Stepwise Regression model (i.e. the traditional approach to identifying significant concurrent predictors in this literature), and compared the amount of variance in Atypicality explained by these models to the amount explained by the PLSR model to ensure that predictive power was maximized. To test our second hypothesis, that discrete linear combinations could likewise be used to predict concurrentBASC-2 Social Skills, we repeated this process. Finally, we conducted post-hoc analyses to examine whether the derived loading patterns were optimally effective for each of three subgroups (Learning Disorders, ADHD, Autism Spectrum Disorders) of participants. We did this by creating dummy coded variables for each subgroup. Then, we specified an OLS regression with three predictors: the subgroup indicator, each child’s score for the specified PLSR component, and the interaction between the two (if there was more than one PLSR component, this process was repeated for each component). If the interaction term was significant (p < .05), it was taken to indicate that the relation between the derived PLSR component and the BASC-2 outcome was especially descriptive of the given subgroup. We note that the present sample size (n = 45) is sufficient to conduct PLSR [64]. Indeed PLSR has been effectively used with samples as small as 10 cases [65], and PLSR often produces greater statistical power than traditional multiple regression [63]. A past Monte Carlo simulation indicates that N~50 is sufficient to model at least 2–3 latent factors [65]. Additionally, PLSR statistical power is affected by normality of the distribution of observed variables [80]; thus, the Z-score transformation discussed above serves to increase statistical power with the present data.
rdf:type