The examination of various socio-economic factors in facilitating the ability to predict methamphetamine lab locations began with the Ordinary Least Squares (OLS) regression to help assure the creation of a good model. Initially all independent (explanatory) variables under consideration in this study were applied to the dependent variable (meth lab locations) as a group. The results were examined for redundancy, statistical significance, and a correlation between the independent and dependent variables. Following this critique of the results, the variable was evaluated for consideration of inclusion based on considered importance to the study. The goal of the OLS regression was a return of a high Adjusted R-Square value. Independent variables were removed individually in an effort to maximize the Adjusted R-Square value based on a critique of the results. This was followed by another run of the OLS tool. This process was reiterated numerous times until satisfactory results were achieved. A table of the final OLS results follows:
To inspect for bias, the OLS regression's Jarque-Bera Statistic Score was checked along with the histograms of a scatter plot matrix. Finally, the Standard Residual values for each census tract were added to the map for visual critique. Most of the two-county area was predicted accurately with this model as shown in the map below. Additional work would include resolving the high and low residuals.
 |
| Final OLS Results Table |
 |
| Meth Lab Locations and Standard Residual Values Charleston, West Virginia, with Kanawha and Putnam Counties |
No comments:
Post a Comment