Significant Factor Score Variability and the Validity …

1 Downloaded By: [CDL Journals Account] At: 19:50 8 July 2008 Significant Factor Score Variability and the Validityof the WISC-III Full scale IQ in PredictingLater Academic AchievementMiranda E. Freberg and Beverly J. VandiverThe Pennsylvania State University, University Park, PennsylvaniaMarley W. WatkinsArizona State University, Tempe, ArizonaGary L. CanivezEastern Illinois University, Charleston, IllinoisThe purpose of this study was to investigate the Validity of the WISC-III (Wechsler,1991) Full scale IQ (FSIQ) scores in predicting later academic achievement given signifi-cant Variability among any of the four WISC-III Factor scores. Taken from an archivaldata set, the sample was composed of 6- to 13-year-old students who were twice evalu-ated for special education eligibility over approximately a three-year retest were separated into two groups based on the presence or absence of signifi-cant Factor Score Variability and then matched across groups on disability, FSIQ, age,sex, and ethnicity.

2 The results of hierarchical multiple regression analyses indicated thatthe FSIQ was a valid predictor of academic achievement scores even in the presence ofsignificant Factor Score words: academic achievement, Factor Score Variability , intelligence, test interpretationResearchers and practitioners have differing opinionsabout which IQ scores should be interpreted on intelli-gence tests as well as which IQ Score global, Factor ,or subtest scores most reliably and accurately predictsacademic achievement (Sattler, 2001). According toSattler, psychologists generally agree that the globalability Score is the most parsimonious, valid predictorof academic achievement. However, some practitionershave contended that Factor and subtest scores provideadditional clinically useful information, including profilesof individuals specific strengths and weaknesses (Donders,1996; Kaufman, 1994; Kaufman & Lichtenberger, 2000).

3 One survey of 354 nationally certified school psychologistsrevealed that 89%of respondents interpreted Factor andsubtest scores to gain additional information above andbeyond the global IQ (Pfeiffer, Reddy, Kletzel, Schmelzer,& Boyer, 2000).Despite the popularity of subtest Score analysis, cur-rent research fails to support this practice (McDermott& Glutting, 1997; Watkins & Canivez, 2004; Watkins& Kush, 1994). Nevertheless, the debate about whichscores to interpret continues, particularly in the predic-tion of academic achievement scores. The currentconsensus is that the global IQ is the best predictor ofacademic achievement in the absence of underlyingdiscrepancies among Factor and subtest scores (Sattler,2001).

4 However, scholars disagree about which IQscores to interpret when Significant Factor or subtestAddress correspondence to Miranda Freberg, The PennsylvaniaState University, 226 CEDAR Building, University Park, PA : NEUROPSYCHOLOGY,15: 131 139, 2008 Copyright#Taylor & Francis Group, LLCISSN: 0908-4282 print=1532-4826 onlineDOI: By: [CDL Journals Account] At: 19:50 8 July 2008 discrepancies occur. Some experts ( , Kaufman, 1994)argue that in the presence of discrepancy, Factor orsubtest scores are more accurate representations of anindividual s cognitive ability and should be used in theprediction of achievement instead of the global IQ fact, it has been contended that Significant factoror subtest Score discrepancies indicate that the globalIQ does not accurately represent an individual s overallintellectual functioning and its use leads to inaccuratepredictions of achievement (Hale & Fiorello, 2001;Weiss, Saklofske, & Prifitera, 2003).

5 Thus, proponentsof this view have argued against the use of the globalIQ when Factor or subtest Variability exists. Sattler(2001) stated that in the presence of such discrepancy the Full scale IQ may represent a forced average ofrather disparate primary skills (p. 321). Weiss et that the global IQ may not be the best measureof a child s cognitive ability if underlying Factor scoresdiffer. For example, they posited that a 37-point discrep-ancy between Factor scores on the Wechsler IntelligenceScale for Children Third Edition (WISC-III; Wechsler,1991) may indicate that the overall Full scale IQ (FSIQ)is less meaningful as a summary of an individual s intel-lectual ability. Under these circumstances, Weiss et that the FSIQ not be reported or onlyreported with caution.

6 Although accepted by manyexperts ( , Kaufman, 1994; Sattler, 2001), this assump-tion has not been extensively empirically FOR THE GLOBAL IQ SCORERyan, Kreiner, and Burton (2002) examined whether sig-nificant intersubtest scatter on the Wechsler Adult Intel-ligence scale Third Edition (WAIS-III; Wechsler, 1997a)compromised the predictive Validity of the FSIQ inpredicting the eight indexes on the Wechsler MemoryScale Third Edition (WMS-III; Wechsler, 1997b) in asample of veteran men suffering from a variety ofmedical and psychiatric disorders with high versus lowintersubtest scatter. High intersubtest scatter was quanti-fied as a nine-point difference or greater between thehighest and lowest subtest scaled scores whereas lowintersubtest scatter was set at less than a nine-point differ-ence.

7 The two groups were matched on FSIQ within onepoint, with no Significant differences found between thetwo groups on FSIQ, age, years of education, distributionof ethnicities, or diagnoses. Separate regression analysesfor the low- and high-scatter groups were used to predictthe WMS-III indexes from the WAIS-III FSIQs. Usingthe Potthoff method to compare intercepts and slopes,no Significant differences were found between the low-and high-scatter groups. That is, the FSIQ was foundto be a valid predictor of WMS-III index scores even inthe presence of intersubtest scatter (Ryan et al., 2002).Kahana, Youngstrom, and Glutting (2002) usedscores obtained on the Differential Ability Scales(DAS; Elliott, 1990) from a national representativesample of 6- to 17-year-old youth to examine (a) theoverall frequency of Significant discrepancies betweenfactor and subtest scores, and (b) the predictive utilityof the Factor and subtest discrepancies in forecastingacademic achievement.

8 Multiple regression analyseswere used to test the incremental Validity of the factorand subtest scores. The General Conceptual Ability(GCA) was entered as the first block for all threeachievement criteria. The second block consisted of thefactor scores, a discrepancy status variable indicatingthe presence or absence of a Significant discrepancybetween the Factor scores, and the interaction of thefactor Score and the discrepancy status variable. Thethird block of each regression model contained the sub-test scores, the Significant subtest discrepancies, and theinteraction between subtest and discrepancy (Kahanaet al., 2002). Results indicated that the GCA was theonly variable in all of the models tested that contributeda Significant amount of unique variance in predictingacademic achievement.

9 The discrepancy and interactionvariables contributed minimal additional variance (1 ) to the prediction. Kahana et al. concluded thatthe GCA was the most parsimonious and robust pre-dictor of all three achievement criteria measured bythe , Youngstrom, Ward, Ward, and Hale (1997)investigated whether the Verbal IQ (VIQ), the Perfor-mance IQ (PIQ), or any of the four Factor scores (VerbalComprehension, Perceptual Organization, Freedomfrom Distractibility, and Processing Speed) on theWISC-III improved the prediction of academic achieve-ment above and beyond the FSIQ Score among a nation-ally representative sample of 283 nonreferred childrenand a sample of 636 children referred for an series of hierarchical regressions was conducted todetermine the relative contributions of the FSIQ, theVIQ, the PIQ, and the four underlying Factor scores inpredicting achievement (Reading, Math, Writing, andLanguage), as measured by the Wechsler IndividualAchievement Test (WIAT; Wechsler, 1992).

10 FSIQ wasalways entered into the regression equation first, fol-lowed either by the four Factor scores or by the VIQand PIQ. In every regression equation, the FSIQ con-tributed most substantially to the prediction of academicachievement. As a group, the four Factor scores accountedfor an additional 5 16%of the variance beyond whatthe FSIQ contributed to the prediction of the achievementcriteria. Separately each Factor Score uniquely contributed0 5%of the variance. Together, the VIQ and PIQaccounted for the variance above and beyondthe FSIQ, and neither the VIQ nor the PIQ accounted forany Significant unique variance. The researchers concluded132 FREBERG, VANDIVER, WATKINS, & CANIVEZD ownloaded By: [CDL Journals Account] At: 19:50 8 July 2008 that the FSIQ was the most parsimonious and powerfulpredictor of academic achievement and that Factor scoresproduced negligible increases in the prediction of allachievement FOR Factor Score ANALYSISHale, Fiorello, Kavanagh, Hoeppner, and Gaither (2001)contended that Glutting et al.