Testing the predictive validity of theories in health behavior research is
highly salient to the development of an evidence base of potentially
modifiable health behavior determinants and the processes by which
they relate to health behavior. Such evidence may signal potential
targets for efficacious behavior change interventions to promote health.
Such tests typically involve evaluating the fit of models specifying theory
stipulated nomological networks among psychological constructs and
outcomes with observed data in health behavior contexts. Analytic
techniques combining meta-analysis and multivariate analyses such as
structural equation modeling affords researchers opportunity to conduct
large-sample tests of such theoretical networks using synthesized data
across multiple studies. Importantly, the techniques also provide
opportunity to test predictions of unique theoretical perspectives,
including integrated theoretical models, that draw constructs and
process-related predictions from multiple perspectives to arrive at
comprehensive accounts of health behavior. Such models, for example,
incorporate multiple processes beyond those specified in typical social
cognition and motivational theories, such as volitional and implicit
processes. In this paper, we provide a non-technical overview of this
approach and outline its advantages including: leveraging of the large
sample size from multiple data sources to provide robust point and
variability estimates of theory effects with high statistical power; capacity
to provide overall evaluations of existing and new theoretical models
across extant research and guideline ranges for theory effects in
subsequent tests; and ability to test mechanistic effects, namely,
mediation effects and effects of study- and sample-level moderator
variables on theory effects. To illustrate our points, we outline a series of
examples that illustrate the application of this approach to test existing
and novel integrated theoretical models in health behavior research. We
conclude by summarizing some of the limitations of this approach and
recommend future innovations that may advance theoretical knowledge
in the field.