Tests of measurement invariance of three Wechsler intelligence tests in economically developing nations in South Asia and Sub-Saharan Africa

ABSTRACT

Tests of measurement invariance are essential to determining whether individual scores or group averages are comparable across populations. While international comparisons of mean IQ scores are common, tests of measurement invariance for intelligence test batteries (necessary for comparisons to be empirically supported) are rare. In this study, four archival sets of Wechsler test IQ scores from Ghana, Kenya, Pakistan, and Sudan were tested for measurement invariance when compared to the American norm data for the same tests. Results indicate that two datasets – from Ghana and Kenya – demonstrated strict measurement invariance. However, the other two data sets failed to meet the requirements of scalar or strict measurement invariance, which indicates that global IQ scores from these latter data sets cannot be compared to American IQ scores on the same tests. Tests of measurement invariance should be regularly conducted when making comparisons of scores between industrialized and economically developing nations.

KEYWORDS:

• Africa
• Asia
• intelligence
• measurement invariance
• Wechsler tests
• intelligence tests
• cross-cultural psychology

Author Note

The author is now an independent scholar who is no longer affiliated with Utah Valley University.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Declaration

I have no conflicts of interest to disclose. This research project received no funding. I appreciate Brian Stone for providing a WISC-III manual for this study, Ross Larsen for some computation assistance, and Zachary Bergson for sharing WAIS-IV raw score means.

Notes

¹ Note that Warne and Burningham’s (2019) study includes summary data were available from a sample of Sudanese children with local IQs ≥ 130 (Attallah et al., 2014). However, because these sample members were selected on the basis of an IQ score, their covariance matrix will not conform to the covariance matrix of the general examinee population (Muthén, 1989). This renders a test of invariance unnecessary because the American and the selected Sudanese groups will not have the same factor structures. This expectation was verified empirically (see supplemental data files at https://osf.io/5ufnt/?view_only=fb600bcbee81493880ee05ded60f8373).

² Bakhiet et al. (2017) did not specify the qualifications that the students must meet to apply to the schools for gifted children in Khartoum. Presumably, their academic achievement and/or intelligence was higher than average.

³ Each procedure was only used with half of the sample data.

⁴ Miezah (2015) also performed an exploratory factor analysis (with an orthogonal varimax rotation) in which a four-factor solution was imposed on the data. This was an attempt to determine whether the Ghanian data conformed to the factor structure reported in the WAIS-IV manual. Unfortunately, this was not an appropriate methodology because the manual’s factor structure was tested with confirmatory factor analysis; as stated later in this article, exploratory factor analysis is a different technique used to answer different empirical questions. The results from the two types of factor analyses are not interchangeable, nor should one expect them to be.

⁵ The three-factor model clearly failed to show scalar invariance: χ² = 568.788, df = 37, p < .001, CFI = .825, RMSEA = 0.133 [90% CI = .123, .142], SRMR = .112.

⁶ Mean scores for this group correspond to an IQ of 87.8 (Gichuhi, 1999, p. 44).

⁷ Mean scores for this group correspond to an IQ of 83 if calculating the weighted average of the sum of subtest scores and then converting this to an IQ. If the groups’ raw score means are converted to IQs first, then the weighted average is 86.4. Note that Miezah (2015, p. 67) reported a mean IQ of 93.77. This is clearly an error, assuming that the raw scores reported in his Table 4 are correct.

⁸ The author attempted to contact Bakhiet to determine whether this was the case, but there was no response, perhaps due to the present unrest in Sudan.

⁹ The Pakistani scaled subtest scores had standard deviations that were 11.0–74.3% lower than the standard deviations for the American norm sample. The verbal IQ, performance, IQ, and full scale IQ all also had lower standard deviations in the Rasheed et al. (2018) sample: 33.6%, 37.7%, and 49.6% lower, respectively.

Additional information

Notes on contributors

Russell T. Warne

Dr. Russell T. Warne earned his PhD in educational psychology (with an emphasis in research, measurement, and statistics) from Texas A&M University. Currently, he is a freelance researcher who specializes in research on intellectual giftedness, human intelligence, and differential psychology. He is the author of the acclaimed book In the Know: Debunking 35 Myths About Human Intelligence, and the undergraduate textbook Statistics for the Social Sciences: A General Linear Model Approach, both published by Cambridge University Press. He has also published 68 articles in peer-reviewed journals.