Analysis

Multiple logistic regression was used to examine the odds of past-year heroin use, NUPO, and concurrent NUPO and heroin use among respondents involved in the sixteen different types of sports when controlling for potentially confounding factors. STATA 14.0 was the software used to estimate the models outlined above; all analyses used the weights provided by the MTF to account for the probability of selection into the sample. Given missing data within the MTF sample, multiple imputation was used to impute missing observations (see Table 1). Finally, design effects reflecting the cluster sampling were not available from the public use MTF study. Accordingly, results at the .01 alpha level or lower were given serious consideration.²