TABLE 1—
Bayesian Poisson Space-Time Regression of Incidence of Methamphetamine Abuse and Dependence: California, 1995–2008
| Parameter | Mean (SD) | MC Error | Median (95% Credible Interval) | Relative Rate |
| Demographic characteristics | ||||
| Average household size | −0.3686 (0.0178) | 0.0009 | −0.3661 (−0.4067, −0.3382)a | 0.6917 |
| Household income × 1000 | −0.0144 (0.0004) | 0.0000 | −0.0144 (−0.0151, −0.0136)a | 0.9857 |
| Proportion White | 0.3129 (0.0433) | 0.0021 | 0.3118 (0.2340, 0.3983)a | 1.3674 |
| Proportion Hispanic | 0.8751 (0.0485) | 0.0023 | 0.8750 (0.7839, 0.9725)a | 2.3991 |
| Population density quintile 2 | 0.0474 (0.0243) | 0.0010 | 0.0475 (0.0007, 0.0958)a | 1.0486 |
| Population density quintile 3 | 0.0966 (0.0265) | 0.0012 | 0.0960 (0.0468, 0.1513)a | 1.1015 |
| Population density quintile 4 | 0.0870 (0.0286) | 0.0013 | 0.0860 (0.0326, 0.1460)a | 1.0909 |
| Population density quintile 5 | 0.0531 (0.0315) | 0.0015 | 0.0525 (−0.0089, 0.1194) | 1.0545 |
| Environmental characteristic: highway class 1 or 2 | 0.0479 (0.0117) | 0.0005 | 0.0477 (0.0256, 0.0719)a | 1.0490 |
| County-level covariates | ||||
| Dangerous drug arrests/100 | 0.4113 (0.1986) | 0.0098 | 0.4180 (0.0139, 0.7863)a | 1.5088 |
| Meth manufacture arrests/100 | −0.9595 (1.1170) | 0.0526 | −0.9479 (−3.2150, 1.2140) | 0.3831 |
| % unemployed | −0.0200 (0.0076) | 0.0004 | −0.0195 (−0.0391, −0.0079)a | 0.9802 |
| Control covariate: total hospital discharges | −0.0010 (0.0003) | 0.0000 | −0.0010 (−0.0016, −0.0003)a | 0.9990 |
| Misalignment and interactions | ||||
| % population misaligned | −0.0633 (0.0209) | 0.0010 | −0.0618 (−0.1050, −0.0281)a | 0.9387 |
| Average household size | 0.0020 (0.0049) | 0.0002 | 0.0022 (−0.0080, 0.0110) | 1.0020 |
| Household income × 1000 | 0.0002 (0.0001) | 0.0000 | 0.0002 (0.0001, 0.0003)a | 1.0002 |
| Proportion White | 0.0518 (0.0165) | 0.0008 | 0.0514 (0.0211, 0.0842)a | 1.0532 |
| Proportion Hispanic | 0.0219 (0.0165) | 0.0007 | 0.0219 (−0.0103, 0.0538) | 1.0221 |
| Random effects | ||||
| County | 0.4159 (0.0216) | 0.0007 | 0.4154 (0.3752, 0.4597) | |
| Spatial | 0.6130 (0.0132) | 0.0006 | 0.6128 (0.5889, 0.6402) | |
| Noise | 0.2234 (0.0105) | 0.0005 | 0.2235 (0.2032, 0.2441) | |
| α = spatial/(spatial + noise) | 0.8824 (0.0117) | 0.0005 | 0.8830 (0.8579, 0.9034)a |
Note. MC error = Monte Carlo error. Credible intervals are expressed as log relative rates. Relative rates estimated for the median are shown in the last column to the right. The model included zip code–level and county-level covariates with separate random effects, with the conditional autoregressive process specified at the zip code level. α represents an approximation of the proportion. error attributable to spatial dependence (see Methods). Year-specific fixed-effect intercepts are not shown.
a
Credible interval has < 5% chance of including zero.