When employing unified, quantitative-qualitative methods such as Epistemic Network
Analysis (ENA), the relative frequency of codes and their co-occurrence is of interest.
However, in projects utilizing a large number of codes, if all codes are included,
the interpretation of these models becomes challenging. In this paper, we provide
three potential approaches to code selection. In the theory-based approach, code clustering
and selection was founded on relevant literature or theory. In the insight-based approach,
clusters of codes were defined by the grounded observations of researchers. Lastly,
in the model-based approach, fully inclusive ENA models were generated to select codes
for future models. We illustrated these approaches using data from our ongoing project
that aims to measure the effects of a health education intervention on near-peer educators’
understanding regarding the biopsychosocial model of health. All three approaches
may be useful in guiding code selection for final ENA models or in providing a baseline
for further refinement of model parameters. By outlining these approaches, this work
contributes to discourse on making conscious and transparent decisions regarding ENA
parameterization.
