Validity Investigations for International Versions of the WISC–V: Informing Evidence Based Assessment

 

Submission type

Symposium

Scheduled

Room 115, 10-07-2019, 13:30 - 15:00

Keywords

Factor Structure, Incremental Validity, EFA, CFA, Hierarchical Multiple Regression

Summary of Symposium

Recent publication of the German WISC–V and the WISC–VUK means educational/school and clinical psychologists in Germany, the UK, and Ireland will need to consider emerging empirical evidence for construct validity to provide ethical interpretations.  Independent peer-reviewed research to date regarding the factor structure of the WISC–V, French WISC–V, Canadian WISC–V, WISC–VUK, and WISC–V Spain have indicated a lack of support for the publisher promoted models and interpretive emphases.  This symposium provides independent assessments of the German WISC–V and WISC–VUK, assisting users in determining adequacy of scores and interpretation practices.

Auteurs

Incremental Validity of the WISC–VUK Factor Index Scores with a Referred Irish Sample

James, K (Kate)¹, Good, R (Rebecca)¹, McGill, RJ (Ryan)², Canivez, GL (Gary)^3
1Eirim: The National Assessment Agency Limited, 16 Greenmount House, D6WV659, Dublin, Ireland
²William & Mary, P. O. Box 8795, 23187, Williamsburg, VA
³Eastern Illinois University, 600 Lincoln Avenue, 61920, Charleston, IL

 

Abstract ID

Submission type

Oral only

Keywords

WISC-V UK, Incremental Validity, Hierarchical Multiple Regression, Validity

Summary

Hierarchical multiple regression assessed proportions of WIAT–IIIUKachievement test variance predicted by the WISC–VUKFSIQ and factor index scores to assess incremental validity (Hunsley,2003; Hunsley & Meyer,2003).  If primary interpretation of the five factor index scores is to be followed, the index scores mustdemonstrate meaningful incremental predictive validitybeyond the FSIQ.  The FSIQ was singularly entered into the first block and the five factor index scores were jointly entered into the second block.  Preliminary analyses of 144 Irish children referred for evaluation of learning difficulties indicated the FSIQ accounted for statistically significant (p<.0001) and medium to large portions of WIAT–IIIUK subtest score variance (12.7%–60.4%). Statistically significant (p<.05) portions of WIAT–IIIUKsubtest score variance was incrementally accounted for by combined WISC–VUKfactor index scores with medium effect sizes for all except Numeracy and Math Problem Solving scores (small effect sizes), ranging from 3.1%–14.5%.  Uniquecontributions of WISC–VUKfactor index scores in separatelypredicting the WIAT–IIIUKsubtests (based on squared part correlations from the predictor entered last in the block entry procedure) were as follows: VCI (0%–0.1%), VSI (0%), FRI (0%), WMI (0.0%–0.1%), PSI (0%).  Interpretation implications will be discussed.

Auteurs

Confirmatory Factor Analyses of the WISC–VUK Primary Subtests with a Referred Irish Sample

Good, R (Rebecca)¹, James, K (Kate)¹, McGill, RJ (Ryan)², Canivez, GL (Gary)^3
1Eirim: The National Assessment Agency Limited, 16 Greenmount House, D6WV659, Dublin, Ireland
²William & Mary, P. O. Box 8795, 23187, Williamsburg, VA
³Eastern Illinois University, 600 Lincoln Avenue, 61920, Charleston, IL

 

Abstract ID

Submission type

Oral only

Keywords

WISC-V UK, CFA, Bifactor Model, Higher-order Model, Validity

Summary

The WISC–VUKAdministration and Scoring Manual presents results from confirmatory factor analyses (CFA) of the UK standardization sample that used the same problematic methods applied to the US WISC–V (Beaujean, 2016; Canivez, Watkins, & Dombrowski, 2016a, 2016b; Canivez & Watkins, 2016).  Details of CFA procedures with the WISC–VUKare lacking (Boomsma, 2000) as are examinations of bifactor structure as an alternative to the higher-order model promoted by the publisher.  Canivez, Watkins, and McGill (2018) reported the best representation of WISC–VUKmeasurement in the standardization sample was a bifactor model with four group factors (VC, PR, WM, and PS) supported by EFA and CFA.  The present paper presents preliminary results from independent WISC–VUKCFA with Irish children referred for evaluation of learning difficulties.  All publisher suggested models with 10 primary subtests were examined with rival bifactor representations to replicate WISC–VUKfindings with a referred sample.  Results replicated independent findings from the WISC–V, Canadian WISC–V, French WISC–V, and WISC–VUK. Model based reliability estimates (Omega-hierarchical) showed adequate portions of true score variance captured by general intelligence but inadequate unique true score variance among the four group factors (VC, PR, WM, PS).

Auteurs

Confirmatory Factor Analyses of the German WISC–V Primary and Secondary Subtests

Buenger, A (Anette)¹, Grieder, S (Silvia)¹, Canivez, GL (Gary)^2
1University of Basel, Missionsstrasse 62, 4055, Basel, Switzerland
²Eastern Illinois University, 600 Lincoln Avenue, 61920, Charleston, IL

 

Abstract ID

Submission type

Oral only

Keywords

German WISC-V, CFA, Bifactor Model, Higher-order Model, Validity

Summary

Recent publication of the German WISC–V (Wechsler, 2017a) shows a parallel instrument to the US version with claims of measuring general intelligence and five group factors (Verbal Comprehension[VC], Visual Spatial[VS], Fluid Reasoning[FR], Working Memory[WM], Processing Speed[PS]). The German WISC–V Technical Manual(Wechsler, 2017b) CFAs were not adequately informative (Boomsma, 2000) as no explanation for choice of estimator was provided (perhaps weighted least squares was used like other WISC–V versions) and there exists numerous threats to model viability including a general intelligence and Fluid Reasoning standardized path of 1.0, cross–loading Arithmetic on three factors (with meaningless VC–AR standardized coefficient of .02), failure to test rival bifactor models, and failure to provide decomposed variance estimates.  This paper presents results from independent German WISC–V confirmatory factor analyses using maximum likelihood estimation.  Models with four or five group factors all fit well, but the best model was the bifactor model with four group factors (identical to EFA results), due to local misfit.  The omega–hierarchical coefficient supported interpretability of a unit-weighted g–factor composite, but small variance portions and low group factor omega–hierarchical subscale coefficients (except PS) indicated unit-weighted group factor composite scores are likely not interpretable.  

Auteurs

Hierarchical Exploratory Factor Analysis of the German WISC–V Primary and Secondary Subtests

Grieder, S (Silvia)¹, Buenger, A (Anette)¹, Canivez, GL (Gary)^2
1University of Basel, Missionsstrasse 62, 4055, Basel, Switzerland
²Eastern Illinois Universtiy, 600 Lincoln Avenue, 61920, Charleston, IL

 

Abstract ID

Submission type

Oral only

Keywords

German WISC-V, Hierarchical EFA, Schmid-Leiman Transformation, Validity

Summary

Recent publication of the German WISC–V (Wechsler, 2017a) shows a parallel instrument to the US version with claims of measuring general intelligence and five group factors (Verbal Comprehension[VC], Visual Spatial[VS], Fluid Reasoning[FR], Working Memory[WM], Processing Speed[PS]).  The German WISC–VTechnical Manual(Wechsler, 2017b) CFAs were not adequately informative (Boomsma, 2000) as no explanation for choice of estimator was provided (perhaps weighted least squares was used like other WISC–V versions) and there exists numerous threats to model viability including a general intelligence and Fluid Reasoning standardized path of 1.0, cross–loading Arithmetic on three factors (with meaningless VC–AR standardized coefficient of .02), failure to test rival bifactor models, and failure to provide decomposed variance estimates.  This paper reports on the first- and second-order German WISC–V exploratory factor analyses with Schmid-Leiman variance partitioning (SL; Schmid & Leiman, 1957) procedure to apportion variance to general intelligence and first-order factors.  Four, not five, first-order factors were adequately defined (VS and FR merged), but most German WISC–V variance is associated with general intelligence, not the group factors.  Model based reliability estimates (omega-hierarchical) showed adequate portions of general intelligence true score variance but inadequate unique group factor true score variance.  

Auteurs