Figure 1. Mean Estimated Impacts from Causal Tree and Random Forest Estimation Strategies.
In this illustration, the blue vertical line is the global estimate of impact from the random forest.
The range of uncertainty is illustrated using a histogram, indicating the proportion of simulations which result in positive or negative estimates.
Specific confidence intervals can be calculated for this global estimate, or for individual project locations. A map of project location uncertainties is presented below.
Propensity Model: Programmatic w/ Bio (Treated), Null Comparison (Control) |
| Dependent variable: |
| |
| treatment |
|
| Dist. to Rivers (m) | 0.0000*** (0.0000, 0.0000) |
| Dist. to Roads (m) | -0.0000* (-0.0000, 0.0000) |
| Elevation (m) | 0.0002* (-0.0000, 0.0005) |
| Slope (degrees) | 0.03 (-0.01, 0.08) |
| Urb. Dist. (rel) | -0.001*** (-0.001, -0.001) |
| Pop. Density (2000) | 0.0003* (-0.0000, 0.001) |
| Protected Area % | 0.004*** (0.003, 0.01) |
| Treecover (2000, %) | 0.002 (-0.005, 0.01) |
| Latitude | 0.04*** (0.03, 0.05) |
| Longitude | 0.01*** (0.01, 0.01) |
| Max Precip. (2002, mm) | 0.0002 (-0.002, 0.002) |
| Min Precip (2002, mm) | -0.01** (-0.02, -0.002) |
| Mean Precip (2002, mm) | 0.002 (-0.003, 0.01) |
| Max Temp (2002, C) | 0.09* (-0.002, 0.19) |
| Min Temp (2002, C) | 0.19*** (0.11, 0.26) |
| Mean Temp (2002, C) | -0.18** (-0.35, -0.02) |
| Nightime Lights (2002, Relative) | 0.01 (-0.01, 0.04) |
| NDVI (2002, Unitless) | -0.0005*** (-0.001, -0.0003) |
| years since implementation | -0.27*** (-0.34, -0.21) |
| Constant | -3.13*** (-4.36, -1.90) |
|
| Observations | 5,028 |
| Akaike Inf. Crit. | 2,437.27 |
|
| Note: | *p<0.1; **p<0.05; ***p<0.01 |
Matched Model: Programmatic w/ Bio (Treated), Null Comparison (Control) |
| Dependent variable: |
| |
| NDVI Diff pre-post implementation |
|
| treatment | -0.01 (-0.07, 0.05) |
| Dist. to Rivers (m) | -0.31 (-0.93, 0.31) |
| Dist. to Roads (m) | 0.07** (0.01, 0.14) |
| Elevation (m) | -0.25*** (-0.36, -0.15) |
| Slope (degrees) | -0.14*** (-0.23, -0.05) |
| Urb. Dist. (rel) | 0.10*** (0.02, 0.17) |
| Pop. Density (2000) | 0.02 (-0.06, 0.11) |
| Protected Area % | -0.03 (-0.09, 0.03) |
| Treecover (2000, %) | 0.16*** (0.06, 0.27) |
| Latitude | -0.07 (-0.25, 0.10) |
| Longitude | -0.23*** (-0.32, -0.14) |
| Max Precip. (2002, mm) | -0.46*** (-0.64, -0.29) |
| Min Precip (2002, mm) | 0.01 (-0.08, 0.10) |
| Mean Precip (2002, mm) | 0.27** (0.06, 0.48) |
| Max Temp (2002, C) | -0.20 (-0.51, 0.11) |
| Min Temp (2002, C) | -0.27 (-0.81, 0.27) |
| Mean Temp (2002, C) | -0.11 (-0.83, 0.61) |
| Nightime Lights (2002, Relative) | -0.002 (-0.08, 0.08) |
| NDVI (2002, Unitless) | -0.07 (-0.16, 0.02) |
| years since implementation | -0.05* (-0.12, 0.01) |
| Urb. Dist. (rel) *Treatment | -0.05 (-0.12, 0.02) |
| Dist. to Rivers (m) *Treatment | 0.25 (-0.37, 0.87) |
| Dist. to Roads (m) *Treatment | -0.02 (-0.09, 0.05) |
| Pop. Density (2000) *Treatment | -0.01 (-0.07, 0.06) |
| Latitude *Treatment | -0.002 (-0.17, 0.17) |
| Longitude *Treatment | 0.005 (-0.09, 0.09) |
| NDVI (2002, Unitless) *Treatment | -0.03 (-0.12, 0.06) |
| Elevation (m) *Treatment | -0.04 (-0.14, 0.07) |
| Slope (degrees) *Treatment | -0.04 (-0.13, 0.06) |
| Treecover (2000, %) *Treatment | 0.12** (0.01, 0.22) |
| Max Temp (2002, C) *Treatment | 0.19 (-0.12, 0.50) |
| Mean Temp (2002, C) *Treatment | -0.40 (-1.12, 0.32) |
| Min Temp (2002, C) *Treatment | 0.18 (-0.36, 0.72) |
| Max Precip. (2002, mm) *Treatment | 0.19** (0.01, 0.36) |
| Mean Precip (2002, mm) *Treatment | -0.17 (-0.38, 0.04) |
| Min Precip (2002, mm) *Treatment | 0.04 (-0.05, 0.13) |
| Protected Area % *Treatment | -0.04 (-0.10, 0.03) |
| years since implementation *Treatment | -0.06* (-0.12, 0.01) |
| Constant | -0.01 (-0.07, 0.05) |
|
| Observations | 870 |
| R2 | 0.33 |
| Adjusted R2 | 0.30 |
|
| Note: | *p<0.1; **p<0.05; ***p<0.01 |