Exploratory and Confirmatory Factor Analysis of the French WISC-V for five age groups

Poster

Reference

Exploratory and Confirmatory Factor Analysis of the French WISC-V for five age groups

LECERF, Thierry, CANIVEZ, Gary

Abstract

The interpretation of the French Wechsler Intelligence Scale for Children-Fifth Edition (WISC-V) is based on a 5-factor model with Verbal Comprehension (VC), Visual Spatial (VS), Fluid Reasoning (FR), Working Memory (WM), and Processing Speed (PS). Evidence of the factorial structure of the French WISC-V was established only through Confirmatory Factor Analyses (CFAs). However, as suggested by several authors, factorial structure should derive from both Exploratory Factor Analysis (EFAs) and CFAs. The first goal of this study was to examine the factorial structure of the French WISC-V with EFAs for five age groups: 6-7, 8-9, 10-11, 12-13, and 14-16 years. For age groups, the WISC-V inter-correlation matrix for the 15 subtests scores was used. Factor extraction criteria (MPA, HPA, BIC, SABIC, etc.) suggested that 1 factor should be retained for age groups 8-9 and 10-11, 2 factors for age group 12-13, and 3 factors for age groups 6-7 and 14-16. None criteria supported a five first-order factors model for these five age groups. To disentangle the contribution of 1st and 2nd order factors, the Schmid and Leiman [...]

LECERF, Thierry, CANIVEZ, Gary. Exploratory and Confirmatory Factor Analysis of the French WISC-V for five age groups. In: 14th European Conference of Psychological Assessment, Lisbonne, Portugal, 5-7 july 2017, 2017

Available at:

http://archive-ouverte.unige.ch/unige:96259

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

1 / 1

Exploratory and Confirmatory Factor Analysis of the French WISC-V for five age groups

T. Lecerf ¹, & G. Canivez ²

* Faculty of Psychology and Educational Sciences, University of Geneva, and University of Lausanne, Switzerland

2 Eastern Illinois University, United States of America

INTRODUCTION

METHOD

OBJECTIVE

CONCLUSIONS RESULTS

14^th European Conference of Psychological Assessment, 2017, Lisbon; [email protected]

q Sample:

§  Publisher refused to provide standardization raw data for

independent analyses so variance-covariance matrices reproduced using correlation matrices and descriptive statistics from manual.

§  The French standardization sample included 1049 children who were divided by age into 11 groups, each consisting of 80 - 104 children.

§  The WISC-V inter-correlation matrices for the 15 subtests scores

were used for 5 age groups: 6-7, 8-9, 10-11, 12-13, and 14-16 years.

q Material:

§  The 15 subtests scores were used (here with the preferred factorial structure proposed by the publisher; with 3 cross-loadings).

q Interpretation of the French Wechsler Intelligence Scale for Children–

Fifth Edition (WISC–V) is based on a 5-factor higher-order model including Verbal Comprehension (VC), Visual Spatial (VS), Fluid

Reasoning (FR), Working Memory (WM), and Processing Speed (PS).

§  In the preferred publisher model, there are three cross-loadings of the Arithmetic subtest on VC, FR, and WM; all other 14 subtests are associated with only one latent variable.

q Evidence of this factorial structure was established through

Confirmatory Factor Analyses (CFA); no Exploratory Factor Analysis (EFA) were reported.

q Publishers indicated that this model was satisfactory for all age

groups (6-7, 8-9, 10-11, 12-13, and 14-16 years); however, no detailed EFAs and CFAs were reported in the French technical manual.

2)  Relationships between latent variables and subtest scores:

§  We used the R-package “OmegaSem,” which do an EFA, take the

highest loading items on each score, and then do a CFA. OmegaSem estimates ω_h, an indicator of the general factor saturation.

1.  Bifactor structure, no specification of the number of factors

2.  Bifactor structure, extraction of 5 factors

3)  Confirmatory factor analyses (model 5e, and 5c)

§  Although global fit indices were adequate for each age group, local fit assessment of the publisher preferred model suggested it was not adequate: Arithmetic does not measure VC.

q  The present study addressed 3 goals:

1. The first goal was to estimate the number of factors in

each age group (6-7, 8-9, 10-11, 12-13, and 14-16 years).

2. The second goal was to estimate the relationship between every latent variable and subtest score, and to estimate

the proportion of variance explained by the general factor.

3. The third goal was to replicate and extend the CFAs for each age group.

1.  Factor extraction criteria suggested to retain only 3 factors. Analyses with OmegaSem support this hypothesis. No EFA evidence for 5 factors was found.

2. The Schmid-Leiman orthogonalization transformation showed that:

•  The g factor was the predominate source of variation among subtests scores;

•  Loadings on the g factor were higher than loading on the specific first-order factors for all subtests scores except for CD, SS, and CA scores.

3.  CFA analyses indicated that:

•  The preferred publisher model with 5-first order factors was not

supported. Higher-order models with 4 first-order factors were more adequate;

•  Bifactor model better fit the data than the higher-order model;

q We recommend users do not discount the Full Scale IQ when

interpreting the index scores of the French WISC-V, because the

general factor accounts for the bulk of the common variance in the

French WISC-V. The 5 index scores conflate broad ability variance with general factor variance, thus are not pure measures of broad abilities.

Evidence shows that French WISC-V index scores among 5 age groups are generally poor indicators of broad abilities when purged of

general intelligence variance thus may be of questionable utility.

VC: Verbal Comprehension

VS: Visual Spatial

FR: Fluid Reasoning

WM: Working Memory

PS: Processing Speed

SI: Similarities BD: Block Design MR: Matrix Reasoning

DS: Digit Span CD: Coding VO: Vocabulary VP: Visual

Puzzles

FW: Figure Weights

PS: Picture Span SS: Symbol Search

IN: Information LN: Letter-

number

CA: Cancellation CO:

Comprehension

AR: Arithmetic AR: Arithmetic AR: Arithmetic

AR: VC-WM-

FR CFI TLI RMSEA SRMR Local fit assessment

6-7 years .938 .917 .069 .050 VC – AR / WM – AR: non significant; g-FR > 1 8-9 years .979 .972 .040 .043 VC – AR / WM – AR: non significant; g-FR > 1 10-11 years .965 .953 .055 .046 VC – AR / FR – AR / W –AR : non significant 12-13 years .915 .885 .088 .057 VC – AR / WM - AR: non significant; FR – AR >

1;

14-16 years .966 .954 .055 .045 VC – AR: non significant; g-Gf > 1 AR: WM-FR CFI TLI RMSEA SRMR Local fit assessment 6-7 years .939 .919 .068 .050 WM – AR: non significant; g-FR > 1 8-9 years .980 .973 .039 .044 WM – AR: non significant; g-FR > 1 10-11 years .966 .955 .054 .045 g-FR > 1

12-13 years .914 .886 .088 .057 WM – AR: non significant; FR – AR > 1; g-FR >

1

14-16 years .966 .955 .054 .045 g-FR > 1

RESULTS

1)  Examination of criteria to determine the number of factors

HPA MAP EV > 1 BIC SABIC SE_scree

6-7 years 3 1 3 2 4 3

8-9 years 4 1 3 3 4 3

10-11 years 3 1 3 3 5 2

12-13 years 3 2 3 3 5 2

14-16 years 4 2 2 2 4 3

ECV of g g loading F1 F2 F3

6-7 years 0.70 .26 (CA) - .76 (AR) SI-VO-IN-CO BD-VP-MR CD-SS-CA 8-9 years 0.71 .33 (CD) - .72 (IN) SI-VO-IN-CO BD-VP-MR CD-SS-CA 10-11 years 0.73 .27 (CA) - .74 (VP) SI-VO-IN-CO BD-VP vs. DS-LN CD-SS-CA-

PS

12-13 years 0.65 .27 (CD) - .85 (VP) SI-VO-IN-CO-DS-LN VP CD-SS-CA- BD 14-16 years 0.70 .27 (CD) - .85 (VP) SI-VO-IN-CO DS-LN CD-SS-CA

ECV of g g loading F1 F2 F3 F4 F5

6-7 years .68 .27 - .74 SI-VO-IN-

CO BD-VP-MR DS CD-SS-CA FW

8-9 years .64 .35 - .70 SI-VO-CO BD-VP-MR-

FW-AR DS-LN-PS CD-SS-CA IN 10-11

years .69 .28 - .75 SI-VO-IN BD-VP AR-DS-LN CD-SS-CA-

PS CO vs. FW 12-13

years NA

14-16

years .65 .47 - .74 SI-VO-IN-

CO BD-VP DS-LN CD-SS-CA PS