Mindful in a random forest: Assessing the validity of mindfulness items using random forests methods

• An statistical method novel to psychometrics was used – random forests.
• Both the one- and the two-dimensional mindfulness instrument showed good predictive quality.
• However, at least half of the items were uninformative.
• It appears that a substantial need for revision of mindfulness items exists.

Whereas the number of studies supporting the efficacy of mindfulness as a health intervention is increasing, the measurement of mindfulness remains a subject of debate. Given the importance of measurement in this field, this paper aims to further our understanding of the assessment of mindfulness by employing an approach referred to as “random forests” (RF). RF is an ensemble learning method that is based on decision trees. RF is well known in biological research, for example, but is practically unknown in psychometrics. In this study, RF was used to gauge the predictive validity of the items from two mindfulness instruments concerning their ability to estimate group allocation (i.e., mindfulness practitioners vs. nonpractitioners). To allow for a better generalization of the results, we examined the research questions in two samples (N = 76 and N = 202) of different quality. We investigated two instruments: the Freiburg Mindfulness Inventory (FMI) and the Mindfulness Attention and Awareness Scale. Although results indicated that both instruments were capable of distinguishing practitioners from nonpractitioners, the predictive quality of most items on both scales was determined to be insufficient.