ABSTRACT
Multinomial logit (MNL) regressions are employed to estimate the likelihood of using one statin over another. Tables with complete results with coefficient estimates, robust standard errors, z-statistics, and confidence intervals appear in Appendix C along with tables for marginal effects calculations. Specific results for explanatory variables of interest are also displayed and interpreted in this section. Sensitivity analyses to test the robustness of the copayment variables, in particular, are included and examined in the following section. MNL regression and marginal effect results were performed using STATA 7. The discussion here will focus primarily on the interpretation of marginal effects estimates.
