Abstract
Genetic algorithms (GAs) are robust machine learning approaches for abbreviating a large set of variables into a shorter subset that maximally captures the variance in the original data. We employed a GA-based method to shorten the 62-item Multidimensional Experiential Avoidance Questionnaire (MEAQ) by half without much loss of information. Experiential avoidance or the tendency to avoid negative internal experiences is a key target of many psychological interventions and its measurement is an important issue in psychology. The 62-item MEAQ has been shown to have good psychometric properties, but its length may limit its use in most practical settings. The recently validated 15-item brief version (BEAQ) is one short alternative, but it reduces the multidimensional scale to a single dimension. We sought to shorten the 62-item MEAQ by half while maintaining fidelity to its six dimensions. In a large nationally representative sample of Americans (N = 7884; 52% female; Age: M = 47.9, SD = 16), we employed a GA method of scale abbreviation implemented in the R package, GAabbreviate. The GA-derived short form, MEAQ-30 with five items per subscale, performed virtually identically to the original 62-item MEAQ in terms of inter-subscales correlations, factor structure, factor correlations, and zero-order correlations and unique latent associations of the six subscales with other measures of mental distress, wellbeing and personal strivings. The two measures also showed similar distributions of means across American census regions. The MEAQ-30 provides a multidimensional assessment of experiential avoidance whilst minimizing participant burden. The study adds to the emerging literature on the utility of machine learning methods in psychometrics.
Highlights
Recent methodological advances in scale abbreviation have demonstrated genetic algorithms (GAs) to be robust machine learning approaches to short-form construction (Yarkoni, 2010; Eisenbarth et al, 2015), which work just as well as traditional approaches (Sandy et al, 2014)
The key goal of our study was methodological: to test how well the GA-derived short form preserved the psychometric properties of the long form Multidimensional Experiential Avoidance Questionnaire (MEAQ)
After trail runs on two imputed datasets used to fine-tune the GA parameters, the following specifications were set for separate 25 GA runs on each of the 25 imputed datasets:
Summary
Recent methodological advances in scale abbreviation have demonstrated genetic algorithms (GAs) to be robust machine learning approaches to short-form construction (Yarkoni, 2010; Eisenbarth et al, 2015), which work just as well as traditional approaches (Sandy et al, 2014). The GAs mimic Darwinian evolution principles to efficienty search for a short form of a long form measure in a fully automated manner (Holland, 1975; Scrucca, 2013). For a long form of length L (e.g., 100 items), the size of the search space for any machine learning method is 2L(1.26e+30) and forms a hypercube of L dimensions. A GA method uses “hypercube sampling” by sampling the corners of the L-dimensional hypercube It optimizes the search for a good solution—the “fittest” short form that maximally explains the variance in the data of the original long-form—by mimicking Darwinian evolution mechanisms of selection, crossover and mutation while searching through a “landscape” of the collection of all possible fitness values to find an optimal value. The technical details of the GA method of scale abbreviation have been described elsewhere (Whitley, 1994; Yarkoni, 2010; Scrucca, 2013)
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
