Table.
Bivariate Model | Multivariate Model | ||||
---|---|---|---|---|---|
Mean (SD) | Coefficient (SE) | P Value | Coefficient (SE) | P value | |
Pseudoephedrine sales, g/100 county residents | 49 (39) | 0.015 (0.003) | <.001 | 0.017 (0.003) | <.001 |
No. of full-time law enforcement officers/100 county residents | 0.12 (0.11) | NA | NA | −1.300 (0.948) | .17 |
Adult population with a high school diploma, % | 0.76 (0.08) | NA | NA | −2.519 (1.375) | .07 |
Abbreviation: NA, not applicable.
Although other models were used to fit the data, including Poisson, zero-inflated Poisson, and negative binomial, the residual distribution indicated an ordinary least-squares linear regression model on the natural log of the number of laboratories in a county as the most appropriate. No pseudoephedrine sales were reported for 4 counties (Bracken, Hancock, Owsley, and Robertson) and they were excluded from the analysis.