Assessment Psychology Online
All about
psychological assessment, testing, and practice management

Assessment Psychology - Solving the Puzzle Assessment Psychology - Looking Behind the Mask Assessment Psychology - Testing and Comparing Assessment Psychology - Examining Individual Differences Assessment Psychology - Exploring the Mind

The clinician's desktop reference for psychological assessment, testing and practice management

   Information resources
for psychologists, mental health practitioners, educators, students and patients

Navigation QuickLinks
Assessment Resources | Practice Resources | Personality | IQ | Genius | Bell Curve | Online Tests | PsychSearch | HealthSearch | Best Psych Sites | Dr Benet | Site Map

Bookmark and Share RSS Feed - AssessmentPsychology.com Site Map - AssessmentPsychology.com

Estimating Premorbid Intelligence
IQ, age, education and sex

 

Regression Equations

Australian Age-Education and Premorbid Cognitive/Intellectual Estimates for the WAIS-III Rebecca Sullivan and Dr. Graeme Senior
Department of Psychology, University of Southern Queensland
and
Dr. Maria Hennessy
Department of Psychology, James Cook University

Poster Presented at the 6th Annual Conference of the
APS College of Clinical Neuropsychologists
Hunter Valley, NSW, Australia
October 12 – October 15, 2000

INTRODUCTION

The third edition of the Wechsler Adult Intelligence Scale (WAIS-III) has been available in Australia for the past three years (Wechsler, 1997). Two limitations in this latest version of one of the most popular cognitive/intellectual test batteries are the absence of normative data permitting adjustment for the impact of different levels of formal education, and equations for estimating premorbid levels of intellectual functioning. This type of information has been available for the WAIS-R since the mid-1980s and has become an integral part of the standard interpretation of this test. Adjustments for the influence of most demographic variables can be performed on individual WAIS-R subtests (Kaufman, McLean, & Reynolds, 1988) and IQ composites (Matarazzo & Herman, 1984; Ryan, Paolo, & Findley, 1991). Additionally, clinicians currently enjoy a number of methods for estimating premorbid intellectual levels using demographic variables (Barona & Chastain, 1986), combinations of demographic variables and WAIS-R subtests (Krull, Scott, & Sherer, 1995), and even estimates based upon other current test performances, such as the NART (Nelson, 1991). The absence of this type of information for the WAIS-III, has led some assessors to caution against the use of the WAIS-III in current clinical practice (Senior, Douglas, & Lange, 1999). These authors recommend continuing to utilise the WAIS-R until these methods become available for the latest edition.

          The role of the current study was to, at least in the Australian context, resolve some of these limitations in the WAIS-III. Since its introduction we have been conducting normative studies designed to examine the relationship of the WAIS-III to other commonly administered psychological tests. While not viable as an Australian standardisation, the data does permit us to formally evaluate the impact of a number of demographic variables upon WAIS-III subtests and composite scores and generate regression equations for estimating WAIS-III composite levels. In addition, methods for estimating WAIS-III composites from other tests such as the NART, the AUSNART, Spot-the-Word, the SILS Vocabulary subtest, and WRAT-3 Reading subtest will be provided.

METHOD

Participants

          The participants in these studies were derived from three normative studies conducted through the Psychology departments at the University of Southern Queensland and James Cook University. In each normative study the relationship between WAIS-III subtests and other tests of cognitive functioning was the primary focus of data collection. Demographic characteristics of the three groups are displayed below.

Table 1
Demographic Characteristics of the Three Normative Groups

 

Sex

Age

Education

 

(Male/Female)

Mean (SD)

Mean (SD)

Group 1 (USQ)

72/57

36.1 (13.3)

12.5 (2.3)

Group 2 (USQ)

48/45

34.2 (8.5)

12.5 (2.7)

Group 3 (JCU)

51/49

42.3 (15.8)

4.6 (1.2)*

* The JCU participants had education coded according to a 6 point coding system where 4 corresponded to 11-12 years of education and 5 corresponded to 13-15 years of education. All USQ data were similarly coded for analyses using combined datasets.

Materials and Procedures

Group 1 participants were administered all WAIS-III verbal subtests, Digit Symbol-Coding, and Symbol Search along with the NART, AUSNART, Shipley Institute of Living Scale, WRAT-3 Reading test, and Spot-the-Word test. Group 2 participants were administered WAIS-III Block Design, Matrix Reasoning and the Working Memory subtests. Group 3 participants were administered all WAIS-III subtests and the NART, AUSNART, Spot-the-Word, and WRAT-3 Reading test. All test administrations were conducted according to standardised instructions found in each test manual.

RESULTS

Influence of Demographic Variables

     Table 2 summarises the findings from individual analyses of variance conducted on each of the WAIS-III subtests and composite scores. The role of these analyses was to examine whether education or sex differentially impacted upon WAIS-III scores when age had already been adjusted for using the tables available in the manual. The table reports the p value for all significant effects in the analyses of variance. The analyses were deliberately conducted on the age-adjusted scaled and standard scores and NOT on raw scores. This is due primarily to the way in which scaled scores, IQ scores, and Factor scores are now computed on the WAIS-III - a clinician attempting to adjust for other demographic variables on the WAIS-III will be doing so with the age-adjusted scores.

As can be seen in the table, performances on virtually all subtests and composites demonstrate significant effects for education, while sex influences only Arithmetic and Digit Symbol-Coding. Surprisingly, despite the use of age-corrected scores a significant main effect was still found for age on the Vocabulary subtest. The significant effects for age on VIQ, FSIQ, and VCI are presumably due to the contribution of the Vocabulary subtest to each of these composites. If this is a robust finding, it raises the concern that the American normative data used to adjust for the influence of age may be insufficient when applied to Australian performances on the Vocabulary subtest. 

With the exception of the Arithmetic subtest and the Full Scale IQ composite no interactions were significant in any of the analyses of variance. This is encouraging as it suggests that the influence of education is essentially additive and would be relatively easy to adjust with appropriate normative data.

Table 2
Summary of Results from  Analyses of Variance of Individual WAIS-III Subtests and Composite Scores

WAIS-III Subtest

N

Age*

Education

Sex

Significant Interactions

Information

129

 

.000

 

 

Vocabulary

338

.004

.000

 

 

Similarities

338

 

.000

 

 

Comprehension

129

 

.003

 

 

Arithmetic

223

 

.001

.035

Sex x Educ x Age (.002)

Digit-Span

223

 

.002

 

 

Letter-Number Seq.

223

 

.001

 

 

Digit Symbol-Coding

129

 

.007

.000

 

Symbol Search

129

 

 

 

 

Matrix Reasoning

93

 

.023

 

 

Block Design

93

 

 

 

 

 

 

 

 

 

 

VIQ

229

.011

.000

 

 

PIQ

100

 

.006

 

 

FSIQ

100

.007

.000

 

Age x Educ (.037)

 

 

 

 

 

 

VCI

229

.002

.000

 

 

POI

100

 

.001

 

 

WMI

229

.03

.002

 

 

PSI

229

 

.013

.001

 

* As all scores analysed in this table were age-adjusted scaled or standard scores no significant effects were anticipated for this variable.

Note:   As of this writing, individual subtest data for the Picture Completion and Picture Arrangement subtests were unavailable for analysis. 

          A second method for investigating the differential impact of age, education, and sex to WAIS-III performance is through examination of raw score variance through stepwise regression. Table 3 displays the results of regressions for each of the subtests where education has been entered first, followed by age, and then finally sex.  In this way the unique contributions of age and sex above and beyond that of education can be examined. Note that these are raw scores and age is now essentially unrestricted.

While it is clear that age accounts for considerable variance in most of the WAIS-III subtests, it is only with the two processing speed measures (Digit Symbol and Symbol Search) that the variance accounted for by age exceeds that accounted for by education. This again stresses the need for WAIS-III normative data that permits clinicians to compensate or account for the contribution of education to WAIS-III subtest performance.   As was the case with its predecessor the WAIS-R, this data would suggest that if one were to choose a single demographic variable for which raw scores should be adjusted it would be education rather than age, at least for the age range of participants in these studies.

Table 3
WAIS-III Subtest Raw Score Variance Accounted for by Education, Age, and Sex.

WAIS-III Subtest
(Raw Scores)

 N

Variance Accounted for by Education

Additional Variance Accounted for by Age

Additional Variance Accounted for by Sex

Information

129

23.3

12.6

1.7

Comprehension

129

13.8

9.3

0.7

Vocabulary

337

13.6

6.7

0.1

Arithmetic

223

11.3

2.9

3.5

Block Design

93

8.9

4.8

4.2

Matrix Reasoning

93

8.8

2.1

0.6

Digit Span

223

4.2

0.9

0.1

Letter-Number Sequencing

223

3.3

0.1

0.0

Digit Symbol-Coding

129

3.0

10.3

12.1

Symbol Search

129

0.9

23.5

0.1

 Predicting WAIS-III Verbal Composites

          One of the most valuable methods of analysis available to neuropsychologists who utilise the WAIS-R as part of their test battery, is the ability to estimate premorbid levels of intellectual functioning. These estimates can then be compared with current performance levels and examined for abnormal discrepancies. The absence of these measures for the WAIS-III seriously limits its clinical utility and the clinician adopting the latest edition of the test is currently giving up some valuable tools with which to test clinical hypotheses.

          The goal of the studies reported below was to correct this problem in Australia by generating regression equations for the WAIS-III. In doing so we were struck with two observations regarding the clinical literature:

1.   While many clinicians and researchers repeatedly highlight the limited utility of IQ scores in favour of the more clinically meaningful factor scores, no studies have generated estimates of WAIS factor indices.

2.   Some of the equations available seem to make little clinical sense, such as estimating Performance IQ based on one’s ability to read irregular words (NART). The majority of measures currently used to estimate level of intellectual functioning are reading tests such as the NART and WRAT-3 Reading subtest, or tests of word knowledge such as the Shipley Institute of Living Scale. It is uncertain how these measures could be expected to predict cognitive abilities such as visual organization.

For these reasons the tables presented below permit the estimation of only WAIS-III Verbal IQ and Verbal Comprehension Index scores. The correlations of the tests administered with PIQ, POI, PSI, and WMI are so small that there would be little clinical value in any equations generated to predict them.

Table 4
Regression Equations for Predicting WAIS-III Verbal IQ from Demographic Variables and a Number of Commonly Administered Verbal Tests

Equation Predicting WAIS-III VIQ

N

r

r2

SEe

85.54 + 5.0(Educ) + 0.2(Age) – 2.87(Sex)

229

.546

.298

9.36

101.32 – 0.58(NARTerr) + 4.18(Educ)

129

.635

.403

8.44

110.51 – 0.48(AUSNARTerr) + 2.97(Educ) – 3.01(Sex)

214

.660

.436

8.50

59.27 + 0.68(STW) + 3.75(Educ) + 0.11(Age) – 2.7(Sex)

229

.604

.364

8.92

45.65+0.98(WRAT3)+ 3.5(Educ)+ 0.11(Age)– 3.24(Sex)

229

.652

.425

8.49

54.90+1.38(SHIPvoc)+3.66(Educ)-0.18(Age)

128

.658

.433

8.28

Table 5
Regression Equations for Predicting WAIS-III Verbal Comprehension Index from Demographic Variables and a Number of Commonly Administered Verbal Tests

Equation Predicting WAIS-III VCI

N

r

r2

SEe

78.88 + 5.32(Educ) + 0.21(Age)

229

.542

.293

9.20

100.28 – 0.54(NARTerr) + 3.75(Educ)

129

.634

.402

8.02

103.30 – 0.46(AUSNARTerr) + 3.36(Educ)

214

.653

.427

8.38

46.71 + 0.85(STW) + 3.74(Educ) +0.10(Age)

229

.632

.400

8.50

45.48 + 0.85(WRAT3) + 3.92(Educ) + 0.13(Age)

229

.626

.392

8.55

57.15 + 1.25(SHIPvoc) + 3.60(Educ) – 0.15(Age)

128

.651

.424

7.92

Where:

      Age = age in years     Sex: 1 = Male     2 = Female

      Educ:                                                    

               1 = less than 9 years                   NARTerr = NART Error Score
               2 = 9 to 10 years                        AUSNARTerr = AUSNART Error Score
               3 = 11 to 12 years                      STW = Spot-the-Word raw score
               4 = 13 to 15 years                      WRAT3 = WRAT3 Reading raw score
               5 = 16 or more years                  SHIPvoc = Shipley Vocabulary raw score

     Each of the equations were generated using hierarchical regression and examined for multicollinearity. In addition to the variables indicated above, an occupational code was also entered into the regression but was never retained in any of the analyses, presumably because it adds no unique variance to the prediction. Because the data in these analyses come from three separate studies, the decision was made not to conduct both a development and cross-validation. However, some comparisons can be made with those equations that correspond to methods developed for the WAIS-R. For example, Barona & Chastain (1986) in predicting WAIS-R VIQ from demographic variables reported a correlation of 0.68 and a standard error of estimate of 10.96. The current study generated a somewhat lower correlation of 0.55 but a smaller standard error of estimate of 9.36. The standard error of estimate for predicting WAIS-R VIQ from NART performance is reported as 7.3 (Nelson, 1991) as compared to 8.44 found for the WAIS-III in the current study.

     Perhaps not surprisingly, all methods of estimation that utilise other test score performances consistently outperform demographic variables alone. Demographic variables account for no more than 30% of WAIS-III VIQ or VCI variance, while only a further 10 to 13% is accounted for by the addition of a current test performance.

DISCUSSION

     Concern has been expressed about the critical need for clinicians to adjust for the impact of education on WAIS-III test performance. In the absence of these corrections, clinicians are likely to over-identify low education individuals as exhibiting deficits on these tests, and under-identify abnormal performances in highly educated individuals. The results of the current study support this concern by demonstrating that substantial variance on virtually all WAIS-III subtests and composites can be accounted for by education. In most cases education accounts for much more variance than does age, the variable that can be currently accommodated through age-adjusted tables. It is beyond the scope of the current data to serve as a mechanism for correcting WAIS-III scores, but they do highlight the critical need for an Australian standardization or, at least, the publication of age-education adjusted scores for the U.S. standardization.    

     This study also provided a series of regression equations to fill the need for methods of estimating WAIS-III composite scores. Due to the verbal nature of the tests employed in the study, equations were generated for predicting only WAIS-III VIQ and VCI scores. Cross-validation of these equations will need to be performed before being adopted into clinical practice.  It is hoped that with the addition of these equations, Australian psychologists will be able to find a greater role for the WAIS-III in their psychological assessments.            

 REFERENCES

Barona, A., & Chastain, R.L. (1986). An improved estimate of premorbid IQ for blacks and whites on the WAIS-R. International Journal of Clinical Neuropsychology, 8 (4), 169-172.

Kaufman, A.S., McLean, J.E., & Reynolds, C.R. (1988). Sex, race, residence, region, and education differences on the 11 WAIS-R subtests. Journal of Clinical Psychology, 44 (2), 231-248.

Krull, K.R., Scott, J.G., & Sherer, M. (1995). Estimation of premorbid intelligence from combined performance and demographic variables. The Clinical Neuropsychologist, 9 (1), 83-88.

Matarazzo, J.D., & Herman, D.O. (1984). Relationship of education and IQ in the WAIS-R standardization sample. Journal of Consulting and Clinical Psychology, 52 (4), 631-640. 

Nelson, H. (1991). National Adult Reading test (NART). Second Edition. Test Manual. NFER-Nelson: Windsor.

Ryan, J.J., Paolo, A.M., & Findley, P.G. (1991). Percentile rank conversion tables for WAIS-R IQs at six educational levels. Journal of Clinical Psychology, 47 (1), 104-107.

Senior, G.J., Douglas, L.A., & Lange, R.T. (1999). Advances in psychological test data interpretation in clinical neuropsychology. Special topic presentation at the 19th Annual Conference of the National Academy of Neuropsychology. San Antonio, Texas, USA. 10-13 November.

 Wechsler, D. (1997). Wechsler Adult Intelligence Scale-Third Edition. San Antonio, Texas: The Psychological Corporation.


About | Contact

Contact Assessment Psychology Online - Please report broken links.

About Assessment Psychology Online

 

Disclaimer | Copyright
Copyright © 2005-2016 - William E. Benet, PhD, PsyD - All Rights Reserved.
Site last updated: July 7, 2016