Are Autistic Traits Measured Equivalently in Individuals With and Without An Autism Spectrum Disorder? An Invariance Analysis of the Autism Spectrum Quotient Short Form

Abstract

It is common to administer measures of autistic traits to those without autism spectrum disorders (ASDs) with, for example, the aim of understanding autistic personality characteristics in non-autistic individuals. Little research has examined the extent to which measures of autistic traits actually measure the same traits in the same way across those with and without an ASD. We addressed this question using a multi-group confirmatory factor invariance analysis of the Autism Quotient Short Form (AQ-S: Hoekstra et al. in J Autism Dev Disord 41(5): 589-596, 2011) across those with (n = 148) and without (n = 168) ASD.Metric variance (equality of factor loadings), but not scalar invariance (equality of thresholds), held suggesting that the AQ-S measures the same latent traits in both groups, but with a bias in the manner in which trait levels are estimated. We, therefore, argue that the AQS can be used to investigate possible causes and consequences of autistic traits in both groups separately, but caution is due when combining or comparing levels of autistic traits across the two groups.

‘Subtypes’ in the Presentation of Autistic Traits in the General Adult Population

“…Similarly, the social subscales of the AQ tended to vary together moderately or strongly, but shared weak negative relationships with the Attention to Detail subscale ( Table 5). The association between the Attention Switching and social domains of the AQ has been reported previously in the literature (Lau et al 2013;Murray et al 2013;Stewart and Austin 2009). This (perhaps counterintuitive) finding is discussed further in the Discussion.…”

The Latent Structure of Autistic Traits: A Taxometric, Latent Class and Latent Profile Analysis of the Adult Autism Spectrum Quotient

“…In many studies AQ score is used as a quantitative variable in order to correlate, predict or otherwise differentiate one group from another on a common scale. Analyses of abbreviated versions of the AQ across individuals with and without ASD suggested caution ought to be taken comparing AQ score across groups as threshold invariance was not observed (Murray et al 2014). A qualification with the findings of the present analysis is that the previous literature has raised concerns about differences between clinical and general samples (Ruzich et al 2015).…”

“…Neither taxometric nor latent class analysis assumes normality so this does not preclude further analysis. Although it is assumed the AQ is normally distributed, the literature reporting distributions for the AQ is mixed, with some finding a normal distribution (Broadbent et al 2013; Hurst et al 2007), some reporting non-normal distributions (Murray et al 2014; Puzzo et al 2009), and many not reporting distributional statistics (Ruzich et al 2015). A histogram of the distribution appears to follow the same broad pattern as reported in the systematic review of AQ data in non-clinical samples (Ruzich et al 2015) (Fig.…”

“…A recent systematic review (Ruzich et al 2015) and psychometric analysis of AQ data (Murray et al 2014) have suggested that taxometric and latent class modelling would be beneficial. Previous taxometric analyses of ASD related constructs have found that a unimodal distribution of observed ATs need not necessarily correspond with a continuous latent structure (Frazier et al 2010).…”

Future Efforts in Flynn Effect Research: Balancing Reductionism with Holism

“…Invariance has been shown to break down when comparing men and women [86]. It has been shown to fail when comparing autistic and non-autistic populations on a test of autistic traits [87]. By extension, it would not be surprising if measurement invariance did not hold when comparing left-handers to right-handers [88], or children of different ages ( [73], p. 457).…”