Brandi A. Weiss

A Comparison of Methods for Estimating Quadratic Effects in Nonlinear Structural Equation Models

Two Monte Carlo simulations were performed to compare methods for estimating and testing hypotheses of quadratic effects in latent variable regression models. The methods considered in the current study were (a) a 2-stage moderated regression approach using latent variable scores, (b) an unconstrained product indicator approach, (c) a latent moderated structural equation method, (d) a fully Bayesian approach, and (e) marginal maximum likelihood estimation. Of the 5 estimation methods, it was found that overall the methods based on maximum likelihood estimation and the Bayesian approach performed best in terms of bias, root-mean-square error, standard error ratios, power, and Type I error control, although key differences were observed. Similarities as well as disparities among methods are highlight and general recommendations articulated. As a point of comparison, all 5 approaches were fit to a reparameterized version of the latent quadratic model to educational reading data.

Latent mean differences in executive function in at-risk preterm children: the delay-deficit dilemma.

OBJECTIVE To examine whether a one-factor executive function (EF) model fit data for three groups of children differing in birth criteria (extremely low birth weight [ELBW], late preterm [LPT], and Term) at each of two chronological ages, 3 and 6 years, and whether the latent mean amount of EF differed. METHODS A retrospective observational cohort study of 1,079 participants; 668 aged 3 years born 2000-2009 (93 ELBW, 398 LPT, and 177 Term) and 411 aged 6 years born 1998-2006 (126 ELBW, 102 LPT, and 183 Term). Latent means analysis was conducted using five indicators for EF: noun fluency, action-verb fluency, similarities reasoning, matrices reasoning, and working memory. RESULTS A one-factor model had acceptable fit for all groups (RMSEA<.06, CFI >0.95, SRMR <0.08). Statistically significant between-groups differences were found for all comparisons except one; there were no statistically significant differences between LPT-Term at age 6. At age 3, ELBW was 0.98 and 1.70 SD below LPT and Term, respectively; LPT was 0.61 SD below Term. At age 6, ELBW was 0.70 and 0.78 SD below LPT and Term, respectively; LPT was 0.10 SD below Term. CONCLUSIONS Executive deficit identified early in development after preterm birth could represent a transient developmental delay likely to resolve at older age or a more subtle adverse effect likely to persist over the life span. Study at multiple age points should assist in resolving this dilemma, which has important implications for early age neuropsychological screening and intervention.

Subtle adverse effects of late preterm birth: a cautionary note.

OBJECTIVE Late preterm birth increases risk of perinatal health complications that typically resolve in the short term. Thus, early elective delivery is thought to have no long-term effects. Whether there is increased risk of adverse psychological outcomes that emerge in early childhood remains uncertain. METHOD The authors compared intellectual, neuropsychological, and behavioral outcomes in 278 late preterm (35-36 weeks) and 192 term (37-41 weeks) participants at age 3 years recruited from a single center, using analysis of variance, analysis of covariance, and regression analyses. Late-preterm participants were further subgrouped by admission to the neonatal intensive care unit (NICU; n = 202) or a well-baby unit (n = 76). Analyses included 132 additional participants born at 34 weeks. RESULTS Late preterm participants had lower general conceptual ability (GCA; i.e., IQ); lower verbal, nonverbal, spatial, visuomotor, and dexterity scores; and poorer adaptability than term participants (p < .01; -0.271 to -0.511 SDs). Gestational age was the most important predictor of these subtle outcomes, not neonatal medical variables; no differences were found between NICU admitted and nonadmitted late-preterm groups. A 1-week increase in gestational age resulted in a 1.941 increase in GCA (d = 0.127). CONCLUSION Gestation is a developmental continuum best not interrupted during its natural course. Our data showing subtle but appreciable effects have important implications for obstetric practice and parental decision making regarding early elective delivery in the absence of maternal or fetal adverse indications.

Developmental normative data for the Baron-Hopkins Board test of spatial location memory

Developmentally appropriate domain-specific tests with strong psychometric properties for preschoolers are lacking and infrequently developed. Baron’s modification of the Hopkins Board test (B-HB) to assess spatial location learning and recall in 3- and 6-year-old children has shown promise in the study of young children born prematurely. Current study data were analyzed on 172 typically developing children at age 3 years and 193 at age 6 years, born at term (≥ 37 weeks; ≥ 2500 grams). Statistically significant gender differences were found and data stratification of T-scores and percentile ranks are provided for each of the eight B-HB measures. The B-HB’s strong interrater reliability (99.5%), low-to-moderate test-retest reliability across the 3-year age span, Pearson correlations showing criterion validity, and differential functioning from other selective attention and visuospatial/visuoperceptual tests provide initial normative data for this novel measure of spatial location memory in young children.

Assessing Spurious Interaction Effects in Structural Equation Modeling: A Cautionary Note

Several studies have stressed the importance of simultaneously estimating interaction and quadratic effects in multiple regression analyses, even if theory only suggests an interaction effect should be present. Specifically, past studies suggested that failing to simultaneously include quadratic effects when testing for interaction effects could result in Type I errors, Type II errors, or misleading interactions. Research investigating this issue has been limited to multiple regression models. Contrarily, structural equation modeling is a more appropriate analysis when hypotheses include latent variables. The current study utilized Monte Carlo simulation to investigate whether quadratic effects should be included in the latent variable interaction model. Consistent with previous research, it was found that including latent variable quadratic effects in the model successfully reduced the frequency of spurious interaction effects but at a cost of low power to detect true interaction effects, inaccurate parameter estimates, inaccurate standard error estimates, and reduced convergence rates. Based on findings from the current study, we recommend that researchers hypothesizing interactions between latent variables should test for these relations using the latent variable interaction model rather than the interaction quadratic model. If researchers are concerned about spurious interactions, then they may want to consider including quadratic effects in the model, provided that they have sample sizes of at least 500 and high indicator reliability. We caution all researchers to base higher order effects models on theory.

An Empirical Evaluation of Mediation Effect Analysis With Manifest and Latent Variables Using Markov Chain Monte Carlo and Alternative Estimation Methods

Recently, the Markov chain Monte Carlo (MCMC) estimation method has become explosively popular in a variety of quantitative research methods. In mediation effect analysis (MEA), the MCMC estimation methods can be a promising tool and an important alternative as compared with traditional methods (e.g., the z test using the delta method and the bias-corrected bootstrapping method) in addressing issues such as nonconvergence and complex modeling. In this article, a subject-level MCMC approach for the single MEA is empirically evaluated and compared with traditional methods through Monte Carlo simulation. The evaluation covers point and interval estimates of both manifest and latent variables across conditions including sample size, effect size, and magnitude of factor loadings. BUGS codes for MEA with both manifest and latent variables are provided that can be easily adapted to fit various MEA models in practice.

The Impact of Parental Support and Perception of School on Hispanic Youth’s Substance Use

With structural equation modeling, the National Survey on Drug Use and Health data were used to examine “parental support” and “perception of school” and their relation to Hispanic youth’s substance use (alcohol, cigarettes, and marijuana).

An Entropy-Based Measure for Assessing Fuzziness in Logistic Regression.

This article introduces an entropy-based measure of data–model fit that can be used to assess the quality of logistic regression models. Entropy has previously been used in mixture-modeling to quantify how well individuals are classified into latent classes. The current study proposes the use of entropy for logistic regression models to quantify the quality of classification and separation of group membership. Entropy complements preexisting measures of data–model fit and provides unique information not contained in other measures. Hypothetical data scenarios, an applied example, and Monte Carlo simulation results are used to demonstrate the application of entropy in logistic regression. Entropy should be used in conjunction with other measures of data–model fit to assess how well logistic regression models classify cases into observed categories.

Test of visuospatial construction: Validity evidence in extremely low birth weight and late preterm children at early school age

The Test of Visuospatial Construction (TVSC), a measure of visuoconstruction that does not rely on upper extremity motor response or written production, was administered to extremely low birth weight (ELBW), late preterm (LPT), and term participants at preschool (n = 355) and kindergarten (n = 265) ages. TVSC showed statistically significant weak-to-moderate positive correlations (age 3: r = .118–.303; age 6: r = .138–.348) with Developmental VMI, Differential Ability Scales-II Copying, Matrices, and Pattern Construction subtests, Baron-Hopkins Board Test, and the Purdue Pegboard. One-way ANOVA indicated ELBW performed worse than Term (p = .044) on visuospatial construction at age 3 with a small-to-medium effect size (d = −0.43). No other statistically significant differences were found at age 3 on the TVSC (ELBW/LPT: p = .608, d = −0.17; LPT/Term: p = .116, d = −0.31). At age 6, ELBW participants performed worse than LPT participants (p = .027) and Term participants (p = .012); LPT participants did not differ from Term participants. Small effect sizes at age 3 (ELBW < LPT, d = −0.17; ELBW < Term, d = −0.43) were notably larger at age 6 (ELBW < LPT, d = −0.42; ELBW < Term, d = −0.53). Important practical differences showing LPT participants performed below Term participants (d = −0.31) at age 3 were no longer evident at age 6 (d = −0.097). These findings provide preliminary evidence of TVSC validity supporting its use to detect neuropsychological impairment and to recommend appropriate interventions in young preterm children.

Entropy-Based Measures for Person Fit in Item Response Theory

This article introduces three new variants of entropy to detect person misfit (Ei, EMi, and EMRi), and provides preliminary evidence that these measures are worthy of further investigation. Previously, entropy has been used as a measure of approximate data–model fit to quantify how well individuals are classified into latent classes, and to quantify the quality of classification and separation between groups in logistic regression models. In the current study, entropy is explored through conceptual examples and Monte Carlo simulation comparing entropy with established measures of person fit in item response theory (IRT) such as lz, lz*, U, and W. Simulation results indicated that EMi and EMRi were successfully able to detect aberrant response patterns when comparing contaminated and uncontaminated subgroups of persons. In addition, EMi and EMRi performed similarly in showing separation between the contaminated and uncontaminated subgroups. However, EMRi may be advantageous over other measures when subtests include a small number of items. EMi and EMRi are recommended for use as approximate person-fit measures for IRT models. These measures of approximate person fit may be useful in making relative judgments about potential persons whose response patterns do not fit the theoretical model.