Model Invariance Across Genders in the Broad Autism Phenotype Questionnaire

Background:

           The ever-increasing prevalence rate of Autism Spectrum Disorders (ASD) necessitates a more thorough understanding of the etiology of the disorder.  ASD is considered one of the most heritable neuropsychiatric disorders; twin and family studies implicate at least moderate genetic heritability in the origin of disorder.  The complexities of ASD, however, impede genetic dissection; researchers have yet to isolate specific autism susceptibility genes. 

            Recent research employs the concept of the Broad Autism Phenotype (BAP), which refers to a milder but similar presentation of traits associated with ASD in relatives of individuals diagnosed with the disorder.  Application of the BAP facilitates the study of genetic effects in the BAP and ASD by allowing for larger sample sizes and isolating specific traits, rather than a cluster of traits in the full condition of ASD, to detect underlying genes.  Several instruments have emerged to capture the BAP.  Empirical evidence suggests that the Broad Autism Phenotype Questionnaire (BAPQ) demonstrates psychometric properties superior to the other self-report BAP measures. 

Objectives:

            The developers of the BAPQ report some strong evidence of reliability, through examination of internal consistency, and validity, based on high sensitivity and specificity of the instrument compared to direct clinical assessment of the BAP.  The evidence of validity presented by the developers, while valuable, is not sufficient for the emerging uses of the BAPQ.  The assumption of model invariance, which refers to the stability of the factor structure of an instrument across groups (e.g., genders), should be upheld in order to draw meaningful conclusions about the results of an instrument.  Recent comparisons of mean differences across gender on the BAPQ assume model invariance without explicitly testing it.  The present study seeks to assess model invariance of the BAPQ across genders, as failing to do so yields serious implications for the interpretation of findings on the measure

Methods:

            The study plans to use Confirmatory Factor Analysis (CFA) to assess model invariance across the groups, as it has emerged as the method of choice.  CFA involves a hierarchy of logically ordered and increasingly restrictive tests of sets of model parameters to investigate the model stability of the BAPQ across genders.  The study intends to test configural, metric, and structural invariance, but the hierarchy depends on the outcomes at each level of analysis.  Model assessment at each level of analysis relies on model fit indices and inferential tests. 

Results:

           The researchers hypothesize that the three-factor structure model of the BAPQ does not vary across groups but complete invariance is not expected.  This study anticipates that configural invariance will be upheld, indicating that the baseline model proposed by the developers of the BAPQ is stable across genders.  The researchers expect to uphold metric invariance, yet will examine partial invariance provided any indications of noninvariance at this level.  The current study does not anticipate stability across groups on structural invariance, given the inherent differences between males and females.

Conclusions:

            Empirical evidence upholding model invariance across gender groups provides invaluable information on the validity of the BAPQ.