Models

Data Sources and Variables
Variable Measure
Outcome variable - from Prison Policy Initiative
Rate of imprisonment per 100,000 residents
Established indicators (demographics and socioeconomic) - from US Census
Black residents percent
Median income in thousands
Residents with at least a bachelor's degree percent
Unemployment rate percent
Rural households percent
Primary explanatory variables - from Substance Abuse and Mental Health Services Administration
Travel time to mental health facility in minutes at 9:00 am on a weekday from the center of the tract
Distance to mental health facility driving miles from the center of the tract
Travel time to inpatient facility in minutes at 9:00 am on a weekday from the center of the tract
Driving miles to inpatient facility driving miles from the center of the tract
Travel time to inpatient facility w/ crisis intervention in minutes at 9:00 am on a weekday from the center of the tract
Driving miles to inpatient facility w/ crisis intervention driving miles from the center of the tract
Reference Data - from American Community Survey
Tract shapfiles geometric vector
\[\begin{aligned} \text{Layer 1:} & Y_{ij} | \mu_j, \sigma_y \sim \text{model of how imprisonment varies WITHIN counties } j \\ \text{Layer 2:} & \mu_j | \mu, \sigma_\mu \sim \text{model of how the typical imprisonment $\mu_j$ varies BETWEEN counties.} \\ \text{Layer 3:} & \mu, \sigma_y, \sigma_\mu \sim \text{prior models for shared global parameters} \\ \\ i = & \text{Tract} \\ j = & \text{County} \\ \end{aligned}\]

Established indicators (demographics and socioeconomic factors)

\[\begin{aligned} Y_{ij} | \beta_{0j}, \beta_1, \beta_2, \beta_3, \beta_4, \beta_5, \sigma_y \sim N(\mu_{ij}, \sigma_y^2) \;\; \text{ with } \;\; \mu_{ij} =& \beta_{0j} + \\ &\beta_1\text{Percent Black}_{ij} + \\ &\beta_2\text{Median Income}_{ij} + \\ &\beta_3\text{Unemployment Rate}_{ij} + \\ &\beta_4\text{Percent with Bachelor's Degree}_{ij} + \\ &\beta_5\text{Urban}_{ij} \\ \\ \beta_{0j} | \beta_0, \sigma_0 \stackrel{ind}{\sim} & N(\beta_0, \sigma_0^2) \\ \beta_{0c} \sim & N(100, 10^2) \\ \beta_1 \sim & N(100, 60) \\ \beta_2 \sim & N(-50, 15) \\ \beta_3 \sim & N(100, 25) \\ \beta_4 \sim & N(-100, 35) \\ \beta_5 \sim & N(100, 25) \\ \sigma \sim & \text{Exp}(l) \\ \sigma_y \sim & \text{Exp}(0.072) \\ \sigma_0 \sim & \text{Exp}(1) \\ \end{aligned}\]

Should I include more factors here? Percent disabled? Percent under poverty with medicaid? Or would that sort of just be the same as median income? Other factors? Maybe geographic mobility or population density or anything else?

Proximity to inpatient/crisis mental health facilities

\[\begin{aligned} Y_{ij} | \beta_{0j}, \beta_6, \sigma_y \sim N(\mu_{ij}, \sigma_y^2) \;\; \text{ with } \;\; \mu_{ij} =& \beta_{0j} + \\ &\beta_6\text{Driving Miles to Inpatient Facility}_{ij} + \\ Y_{ij} | \beta_{0j}, \beta_5, \beta_6, \sigma_y \sim N(\mu_{ij}, \sigma_y^2) \;\; \text{ with } \;\; \mu_{ij} =& \beta_{0j} + \\ &\beta_5\text{Urban}_{ij} \\ &\beta_6\text{Driving Miles to Inpatient Facility}_{ij} \\ \\ Y_{ij} | \beta_{0j}, \beta_1, \beta_2, \beta_3, \beta_4, \beta_5, \beta_6, \sigma_y \sim N(\mu_{ij}, \sigma_y^2) \;\; \text{ with } \;\; \mu_{ij} =& \beta_{0j} + \\ &\beta_1\text{Percent Black}_{ij} + \\ &\beta_2\text{Median Income}_{ij} + \\ &\beta_3\text{Unemployment Rate}_{ij} + \\ &\beta_4\text{Percent with Bachelor's Degree}_{ij} + \\ &\beta_5\text{Urban}_{ij} \\ &\beta_6\text{Driving Miles to Inpatient Facility}_{ij} \\ \\ \beta_{0j} | \beta_0, \sigma_0 \stackrel{ind}{\sim} & N(\beta_0, \sigma_0^2) \\ \beta_{0c} \sim & N(100, 10^2) \\ \beta_1 \sim & N(100, 60) \\ \beta_2 \sim & N(-50, 15) \\ \beta_3 \sim & N(100, 25) \\ \beta_4 \sim & N(-100, 35) \\ \beta_5 \sim & N(100, 25) \\ \beta_6 \sim & N(100, 25) \\ \sigma \sim & \text{Exp}(l) \\ \sigma_y \sim & \text{Exp}(0.072) \\ \sigma_0 \sim & \text{Exp}(1) \\ \end{aligned}\]

Capacity of nearest facilities

\[\begin{aligned} Y_{ij} | \beta_{0j}, \beta_7, \sigma_y \sim N(\mu_{ij}, \sigma_y^2) \;\; \text{ with } \;\; \mu_{ij} =& \beta_{0j} + \\ &\beta_7\text{Capacity of Nearest Inpatient Facility}_{ij} + \\ Y_{ij} | \beta_{0j}, \beta_5, \beta_7, \sigma_y \sim N(\mu_{ij}, \sigma_y^2) \;\; \text{ with } \;\; \mu_{ij} =& \beta_{0j} + \\ &\beta_5\text{Urban}_{ij} \\ &\beta_7\text{Capacity of Nearest Inpatient Facility}_{ij} \\ \\ Y_{ij} | \beta_{0j}, \beta_1, \beta_2, \beta_3, \beta_4, \beta_5, \beta_7, \sigma_y \sim N(\mu_{ij}, \sigma_y^2) \;\; \text{ with } \;\; \mu_{ij} =& \beta_{0j} + \\ &\beta_1\text{Percent Black}_{ij} + \\ &\beta_2\text{Median Income}_{ij} + \\ &\beta_3\text{Unemployment Rate}_{ij} + \\ &\beta_4\text{Percent with Bachelor's Degree}_{ij} + \\ &\beta_5\text{Urban}_{ij} \\ &\beta_7\text{Capacity of Nearest Inpatient Facility}_{ij} \\ \\ \beta_{0j} | \beta_0, \sigma_0 \stackrel{ind}{\sim} & N(\beta_0, \sigma_0^2) \\ \beta_{0c} \sim & N(100, 10^2) \\ \beta_1 \sim & N(100, 60) \\ \beta_2 \sim & N(-50, 15) \\ \beta_3 \sim & N(100, 25) \\ \beta_4 \sim & N(-100, 35) \\ \beta_5 \sim & N(100, 25) \\ \beta_7 \sim & N(100, 25) \\ \sigma \sim & \text{Exp}(l) \\ \sigma_y \sim & \text{Exp}(0.072) \\ \sigma_0 \sim & \text{Exp}(1) \\ \end{aligned}\]

Both (Interaction?)

\[\begin{aligned} Y_{ij} | \beta_{0j}, \beta_6, \sigma_y \sim N(\mu_{ij}, \sigma_y^2) \;\; \text{ with } \;\; \mu_{ij} =& \beta_{0j} + \\ &\beta_6\text{Driving Miles to Inpatient Facility}_{ij} + \\ &\beta_7\text{Capacity of Nearest Inpatient Facility}_{ij} + \\ Y_{ij} | \beta_{0j}, \beta_5, \beta_6, \sigma_y \sim N(\mu_{ij}, \sigma_y^2) \;\; \text{ with } \;\; \mu_{ij} =& \beta_{0j} + \\ &\beta_5\text{Urban}_{ij} \\ &\beta_6\text{Driving Miles to Inpatient Facility}_{ij} + \\ &\beta_7\text{Capacity of Nearest Inpatient Facility}_{ij} \\ \\ Y_{ij} | \beta_{0j}, \beta_1, \beta_2, \beta_3, \beta_4, \beta_5, \beta_6, \sigma_y \sim N(\mu_{ij}, \sigma_y^2) \;\; \text{ with } \;\; \mu_{ij} =& \beta_{0j} + \\ &\beta_1\text{Percent Black}_{ij} + \\ &\beta_2\text{Median Income}_{ij} + \\ &\beta_3\text{Unemployment Rate}_{ij} + \\ &\beta_4\text{Percent with Bachelor's Degree}_{ij} + \\ &\beta_5\text{Urban}_{ij} \\ &\beta_6\text{Driving Miles to Inpatient Facility}_{ij} + \\ &\beta_7\text{Capacity of Nearest Inpatient Facility}_{ij} \\ \\ \beta_{0j} | \beta_0, \sigma_0 \stackrel{ind}{\sim} & N(\beta_0, \sigma_0^2) \\ \beta_{0c} \sim & N(100, 10^2) \\ \beta_1 \sim & N(100, 60) \\ \beta_2 \sim & N(-50, 15) \\ \beta_3 \sim & N(100, 25) \\ \beta_4 \sim & N(-100, 35) \\ \beta_5 \sim & N(100, 25) \\ \beta_6 \sim & N(100, 25) \\ \beta_7 \sim & N(100, 25) \\ \sigma \sim & \text{Exp}(l) \\ \sigma_y \sim & \text{Exp}(0.072) \\ \sigma_0 \sim & \text{Exp}(1) \\ \end{aligned}\]