As an initial step, we conducted analyses comparable to those used in the prior report on medical marijuana policy and suicide, employing data from the same time period and including the same set of covariates (Anderson et al., 2014). However, our analyses utilized individual-level data modeled via logistic regression, whereas the previous report described the association between log-transformed state-level suicides rates modeled from aggregated data using linear regression. Because of these differences, we refer to our initial analyses as “comparison analyses” rather than “replication analyses.” In these comparison analyses, we used the same medical marijuana policy coding and the same set of covariates as the previous report: average annual unemployment rate, per-capita income, beer excise taxes, zero-tolerance policies for youth driving under the influence, blood-alcohol content limits of 0.08 for drivers (vs. 0.10), and marijuana decriminalization policy indicators. In the main set of analyses, we extended the observation period from 1990–2007 to 1990–2010; four additional states (Arizona, Michigan, New Jersey, and New Mexico) and the District of Columbia passed medical marijuana policies during this time (Lynne-Landsman et al., 2013). Initial
