Statistical Modeling

The benefit of measuring wellbeing pathways is that university programming staff will have more information with which to create targeted, evidence-based interventions to support wellbeing. Creating such a measure presents two psychometric challenges: (a) it is very large, and (b) statistical analyses of the measure require full structural equation modeling rather than just factor analytic techniques. To manage these challenges, we use planned missing data designs and structural equation modeling methods that are appropriate for data that are both missing and categorical. 

We conduct structural equation and factor modeling using Mplus© software (2017) to determine whether items fit into their assigned places in the Engine Model. Details about our use of structural equation modeling as a method of establishing structural validity (using Messick, 1995) can be found in the attached paper, presented at the annual meeting of The National Council on Measurement in Education (NCME). 

An example of our approach to these structural equation models is shown below in Figure 1. In that figure, the left-most blue box are pathway items; the central red box is outcome items. The right-most green box includes other important outcomes that are associated with wellbeing in the dimension. Examples include: overall life satisfaction, mood, academic engagement, and intent to transfer. 

Figure 1. SEM version of Engine Model for a single dimension


Based on our Spring 2016 survey administration, five of our wellbeing dimensions – meaning, purpose, activity engagement, positivity, and belonging – fit their intended structural models. Using multiple imputation to account for missing data and the WLSMV estimator, these models explain between 55% and 83% of latent factor variance. Using Raykov’s (1997) composite reliability for congeneric measures model (and continuous data analytic methods), these dimensions had factor reliabilities ranging from .855 to .927. Of these dimensions, activity engagement has the strongest associations with our outcomes (affective wellbeing, academic engagement, intent to transfer) followed by meaning.