Construct Validity of WISC-IV: Bayesian Structural Equation Modeling with a Large US Sample

Poster

Reference

Construct Validity of WISC-IV: Bayesian Structural Equation Modeling with a Large US Sample

CANIVEZ, Gary L., et al.

CANIVEZ, Gary L., et al . Construct Validity of WISC-IV: Bayesian Structural Equation Modeling with a Large US Sample. In: 28th International Congress of Applied Psychology , Paris, 8-13 July 2014, 2014

Available at:

http://archive-ouverte.unige.ch/unige:38060

Disclaimer: layout of this document may differ from the published version.

1 / 1

Construct Validity of WISC–IV: Bayesian Structural Equation Modeling with a Large US Sample

Gary L. Canivez ¹ , Philippe Golay ^2,3 , Thierry Lecerf ² , & Marley W. Watkins ⁴

1

Eastern Illinois University,

University of Geneva,

University of Lausanne,

Baylor University

EP020018

g

VCI

PRI

WMI

PSI

Similarities

Comprehension

Block Design

Matrix Reasoning Picture Concept

Digit Span

Letter Number

Coding

Symbol Search Vocabulary

g VCI

PRI

WMI

PSI

Similarities

Comprehension

Block Design

Matrix Reasoning Picture Concept

Digit Span

Letter Number

Coding

Symbol Search Vocabulary WISC-IV

Higher-order model

General intelligence (g) has traditionally been

conceptualized as a superordinate factor (higher-order model) but most recent research has shown better support for g as a breadth factor (bifactor model):

Goal 1: Compare Higher-order (indirect hierarchical) versus Bifactor (direct hierarchical) models of the 10 WISC-IV core subtests from a large referred US sample.

Many controversies remain on the nature of the constructs measured by each subtest score:

Goal 2: Determine more precisely which constructs are adequately measured by WISC-IV core subtests and can secondary interpretation of some subtest scores be supported by the data ?

WISC-IV Bifactor model

Latent variable 1

Subtest Subtest Subtest

Subtest Subtest

Subtest

Informative priors (zero mean and small variance)

Latent variable 2

WISC-IV data were obtained from 1130 US children who were referred for school based clinical evaluation of learning difficulties.

Bayesian estimation combined prior distributions for all parameters with the experimental data and formed posterior distributions via Bayes' theorem.

The prior variance was 0.01 which results in 95% credibility interval of ± 0.20 (small cross-loadings).

BSEM was conducted using Mplus 7.0 with Markov Chain Monte Carlo (MCMC) estimation algorithm with Gibbs sampler. Three chains with 50,000 iterations, different starting values, and different random seeds were estimated.

Sample N % Male Minimum

Age Maximum

Age Mean Age (SD)

US 1130 62%

(696) 6-0 16-11 10.24

(2.51)

BAYESIAN STRUCTURAL EQUATION MODELING (BSEM)

•

BSEM overcomes CFA’s limitations by replacing fixed-to-zero-loadings with “approximate” zeros that translates into small, but not necessarily zero, cross-loadings.

•

Approximate zeros often reflect more theoretically accurate assumptions and facilitate unbiased estimations of the model parameters.

•

Because all relationships between factors and subtest scores are estimated BSEM eliminates the need for comparisons of many competing models. It is also possible to determine the precise nature of the constructs measured by the core subtest scores of the WISC-IV.

Diffuse non informative priors (zero mean and infinite variance)

EP020018

• Results showed that the bifactor model fit was better than the higher-order solution. Models including small cross-loadings were more adequate.

• BSEM suggested a simple and parsimonious interpretation of the subtest scores with the general intelligence factor as a breadth factor directly influencing subtests. This replicates findings from numerous studies across Wechsler versions and other intelligence tests.

• Loadings of the subtests scores on the g-factor were systematically higher than their respective loadings on the four index scores. Index scores represented rather small deviations from unidimensionality and did not necessarily provide sufficient additional and separate information from the general intelligence factor. Omega-hierarchical coefficients for group (index) factors were likely too low for interpretation.

• BSEM allows estimation of models that were closer to theoretical assumptions. BSEM also permits testing more complex models that are not possible to estimate through maximum likelihood estimation.

Model

Posterior Predictive P-

Value

DIC

1. WISC-IV Higher-Order model 0.000 25983.228 2. WISC-IV Higher-Order model with

cross-loadings (priors variance = 0.01) 0.151 25941.978 3. WISC-IV Bifactor model 0.000 25963.109 4. WISC-IV Bifactor model with cross-

loadings (priors variance = 0.01) 0.388 25913.312 Note. Higher Posterior Predictive P-Value and Lower DIC indicates better fit to the data.

EP020018

Results