Beyond Population Level Effects: Addressing Sample Heterogeneity in Prevention Research

Open Access
Abar, Beau
Graduate Program:
Human Development and Family Studies
Doctor of Philosophy
Document Type:
Date of Defense:
May 05, 2010
Committee Members:
  • Eric Loken, Dissertation Advisor
  • Eric Loken, Committee Chair
  • Michael J Rovine, Committee Member
  • Peter Cm Molenaar, Committee Member
  • Robert James Turrisi, Committee Member
  • mixture models
  • prevention
  • skin cancer
The unifying theme of my three-paper dissertation was the examination of interindividual differences using two general methods for illustrating heterogeneity in a sample: mixture models and person-specific analysis. Study 1 examined issues of estimation, classification, and parameter recovery in non-identified latent class models that appear identified (i.e., positive df and model convergence). A model was simulated that was known to be non-identified (4 indicators – 3 classes), and these models were found to often provide interpretable solutions that do not represent the true population model, which resulted in high misclassification of individuals and poor parameter recovery. Study 2 used group-level and person-specific Markov models to examine the pattern of indoor tanning over time, as well as the extent to which current and previous day weather conditions affect this pattern. Results indicated that there was substantial heterogeneity across individuals in patterns of tanning over time. In addition, there appeared to be no relationship between weather conditions and the probability of transitioning to the “Tanning” state from the “No Tanning” state when examined at the group-level and in the majority of person-specific analyses. Study 3 evaluated the efficacy of an intervention aimed at reducing indoor tanning and examined potential heterogeneity in this tanning over time. Results indicated that the intervention was successful at decreasing indoor tanning across the length of the study, and that there were three patterns of tanners among the treatment individuals: abstainers, moderate tanners, and heavy tanners. The intervention was shown to have a harm reduction effect by reducing levels of exposure within the moderate and heavy tanner classes. Substantive and methodological contributions to the field were discussed for each study, as well as for the dissertation as a whole.