Application of Crash Modification Factors for Access Management, Volume 1: Practitioner's Guide (2021)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

156 This appendix provides corridor-level models to support the methods in Chapter 5. The models are organized by land use (mixed-use, commercial, and residential) and crash type (total, injury, turning, rear-end, and right-angle). Note the following when applying the models: • Specific crash types cannot be summed to calculate total crashes. • North Carolina crashes coded as rear-end turn crashes are included in both rear-end and turning crashes. • Models are not provided for property-damage-only crashes because of inconsistent reporting. • Consider the following factors in selecting the preferred model when more than one option is available for the land use and crash type of interest: (1) statistical significance of the coef- ficients for the variables of interest as indicated by the size of the p-value (a lower p-value indicates a higher level of significance) and (2) model fit (a smaller value of k indicates a better-fitting model). Mixed-Use Models Total Crashes Model form: Crashes/mile/year = exp(intercept+Region) ∗ (AADT)b ∗ exp(c1∗X1+ . . . +cn∗Xn) A P P E N D I X E Summary of Models by Land Use and Crash Type Variable Estimate Standard Error P-Value Intercept -3.1845 1.9550 0.1033 Region 1.1410 0.2316 <0.0001 AADT 0.5187 0.1819 0.0043 ACCDENS 0.0053 0.0044 0.2279 SIGDENS 0.1095 0.0607 0.0710 PROPLANE1 -0.5185 0.3789 0.1711 Dispersion (k) 0.5073 -- -- Comments: The p-values for ACCDENS and PROPLANE1 are larger than desirable. Region is included for NC or MN; a value of 0 if in northern or southern CA. ACCDENS = number of driveways plus unsignalized intersections per mile. SIGDENS = number of signalized intersections per mile. PROPLANE1 = proportion of corridor length with two lanes. Table E-1. Alternate Model 1 for mixed-use total crashes.

Summary of Models by Land Use and Crash Type 157   Injury Crashes Model form: Crashes/mile/year = exp(intercept+Region) ∗ (AADT)b ∗ exp(c1∗X1+ . . . +cn∗Xn) Variable Estimate Standard Error P-Value Intercept -3.2905 1.8743 0.0792 Region 1.0533 0.2086 <0.0001 AADT 0.5266 0.1738 0.0024 UNSIGDENS 0.0471 0.0224 0.0354 SIGDENS 0.0957 0.0594 0.1072 PROPLANE1 -0.6376 0.3796 0.0931 Dispersion (k) 0.4897 -- -- Comments: Similar model to Alternate Model 1, but excluding driveways. Region is included for NC or MN; a value of 0 if in northern or southern CA. UNSIGDENS = number of unsignalized intersections per mile. SIGDENS = number of signalized intersections per mile. PROPLANE1 = proportion of corridor length with two lanes. Table E-2. Alternate Model 2 for mixed-use total crashes. Variable Estimate Standard Error P-Value Intercept -0.8926 0.5021 0.0755 Region 0.6166 0.1013 <0.0001 AADT 0.3766 0.0468 <0.0001 PROPNODEV -0.4252 0.2268 0.0608 Dispersion (k) 0.5165 -- -- Comments: This model was developed from the dataset combining all land use types with factor variables representing the land use. Region is included for NC or MN; a value of 0 if in northern or southern CA. PROPNODEV = proportion of length with no roadside development. Table E-3. Alternate Model 3 for mixed-use total crashes. Variable Estimate Standard Error P-Value Intercept -3.5700 1.7816 0.0451 Region 0.5695 0.1980 0.0040 AADT 0.5010 0.1659 0.0025 SIGDENS 0.1239 0.0556 0.0258 PROPLANE1 -0.5814 0.3582 0.1046 Dispersion (k) 0.4248 -- -- Comments: Region is included for NC or MN; a value of 0 if in northern or southern CA. PROPLANE1 = proportion of corridor length with two lanes. SIGDENS = number of signalized intersections per mile. Table E-4. Alternate Model 1 for mixed-use injury crashes.

158 Application of Crash Modification Factors for Access Management Turning Crashes Model form for Alternate Model 1 and Model 2: Crashes/MVMT = exp(intercept+Region) exp(c1∗X1+ . . . +cn∗Xn) Model form for Alternate Model 3: Crashes/mile/year = exp(intercept+Region) ∗ (AADT)b ∗ exp(c1∗X1+ . . . +cn∗Xn) Variable Estimate Standard Error P-Value Intercept -1.7775 0.5964 0.0029 Region 0.2465 0.0931 0.0081 AADT 0.3880 0.0558 <0.0001 PROPNODEV -0.3159 0.2201 0.1511 PROPLANE1 -0.6623 0.1404 <0.0001 Dispersion (k) 0.4151 -- -- Comments: The p-value for PROPNODEV is larger than desirable. This model was developed from the dataset combining all land use types with factor variables representing the land use. Region is included for NC or MN; a value of 0 if in northern or southern CA. PROPLANE1 = proportion of corridor length with two lanes. PROPNODEV = proportion of length with no roadside development. Table E-5. Alternate Model 2 for mixed-use injury crashes. Variable Estimate Standard Error P-Value Intercept -2.1083 0.4338 <0.0001 Region 0.9647 0.2843 0.0007 SIGDENS 0.1865 0.0754 0.0134 ACCDENS 0.0088 0.0061 0.1486 Dispersion (k) 0.7920 -- -- Comments: The p-value for ACCDENS is larger than desirable. Region is included for NC or MN; a value of 0 if in northern or southern CA. ACCDENS = number of driveways plus unsignalized intersections per mile. SIGDENS = number of signalized intersections per mile. Table E-6. Alternate Model 1 for mixed-use turning crashes. Variable Estimate Standard Error P-Value Intercept -2.0792 0.3963 <0.0001 Region 0.8015 0.2354 0.0007 SIGDENS 0.1797 0.0742 0.0154 UNSIGDENS 0.0582 0.0323 0.0719 Dispersion (k) 0.7780 -- -- Comments: Similar model to Alternate Model 1, but excluding driveways. The overall fit of the model improves and the p-value for unsignalized intersections is improved. Region is included for NC or MN; a value of 0 if in northern or southern CA. SIGDENS = number of signalized intersections per mile. UNSIGDENS = number of unsignalized intersections per mile. Table E-7. Alternate Model 2 for mixed-use turning crashes.

Summary of Models by Land Use and Crash Type 159   Rear-End Crashes Model form: Crashes/mile/year = exp(intercept+Region) ∗ (AADT)b ∗ exp(c1∗X1+ . . . +cn∗Xn) Variable Estimate Standard Error P-Value Intercept -0.4146 0.7632 0.5870 Region -0.3163 0.1301 0.0150 AADT 0.2179 0.0729 0.0028 PROPNODEV -0.5890 0.2827 0.0372 Dispersion (k) 0.7791 -- -- Comments: This model was developed from the dataset combining all land use types with factor variables representing the land use. Region is included for NC or MN; a value of 0 if in northern or southern CA. PROPNODEV = proportion of length with no roadside development. Table E-8. Alternate Model 3 for mixed-use turning crashes. Variable Estimate Standard Error P-Value Intercept -6.6976 1.9985 0.0008 Region 1.2289 0.2479 <0.0001 AADT 0.7901 0.1876 <0.0001 SIGDENS 0.1122 0.0702 0.1099 Dispersion (k) 0.7006 -- -- Comments: Region is included for NC or MN; a value of 0 if in northern or southern CA. SIGDENS = number of signalized intersections per mile. Table E-9. Alternate Model 1 for mixed-use rear-end crashes. Variable Estimate Standard Error P-Value Intercept -3.3091 0.6700 <0.0001 Region 0.8113 0.1136 <0.0001 AADT 0.5015 0.0618 <0.0001 SIGDENS 0.0621 0.0380 0.1021 PROPLANE1 -0.5548 0.1713 0.0012 Dispersion (k) 0.6098 -- -- Comments: This model was developed from the dataset combining all land use types with factor variables representing the land use. Region is included for NC or MN; a value of 0 if in northern or southern CA. PROPLANE1 = proportion of corridor length with two lanes. SIGDENS = number of signalized intersections per mile. Table E-10. Alternate Model 2 for mixed-use rear-end crashes.

160 Application of Crash Modification Factors for Access Management Right-Angle Crashes Model form: Crashes/mile/year = exp(intercept+Region) ∗ (AADT)b ∗ exp(c1∗X1+ . . . +cn∗Xn) Variable Estimate Standard Error P-Value Intercept -5.8048 1.9472 0.0029 Region 1.8390 0.2616 <0.0001 AADT 0.4656 0.1856 0.0121 ACCDENS 0.0112 0.0051 0.0267 SIGDENS 0.2284 0.0637 0.0003 Dispersion (k) 0.5585 -- -- Comments: Region is included for NC or MN; a value of 0 if in northern or southern CA. ACCDENS = number of driveways plus unsignalized intersections per mile. SIGDENS = number of signalized intersections per mile. Table E-11. Alternate Model 1 for mixed-use right-angle crashes. Variable Estimate Standard Error P-Value Intercept -5.2671 2.1768 0.0155 Region 1.2134 0.2457 <0.0001 AADT 0.5678 0.2103 0.0069 PROPDIV -0.4710 0.3461 0.1736 MEDOPDENS 0.1901 0.0884 0.0316 Dispersion (k) 0.6796 -- -- Comments: The p-value for PROPDIV is higher than desirable, but the direction of effect for PROPDIV and MEDOPDENS are logical. Region is included for NC or MN; a value of 0 if in northern or southern CA. MEDOPDENS = number of median openings per mile. PROPDIV = proportion of corridor length with divided median. Table E-12. Alternate Model 2 for mixed-use right-angle crashes. Variable Estimate Standard Error P-Value Intercept -2.1485 0.6851 0.0017 Region 1.2344 0.1377 <0.0001 AADT 0.2433 0.0648 0.0002 PROPFULLDEV 0.6787 0.1846 0.0002 Dispersion (k) 0.7674 -- -- Comments: This model was developed from the dataset combining all land use types with factor variables representing the land use. Region is included for NC or MN; a value of 0 if in northern or southern CA. PROPFULLDEV = proportion of corridor length with full roadside development. Table E-13. Alternate Model 3 for mixed-use right-angle crashes.

Summary of Models by Land Use and Crash Type 161   Commercial Models Total Crashes Model form: Crashes/mile/year = exp(intercept+Region) ∗ (AADT)b ∗ exp(c1∗X1+ . . . +cn∗Xn) Variable Estimate Standard Error P-Value Intercept -0.7017 0.6873 0.3073 Region 0.8353 0.1883 <0.0001 AADT 0.3094 0.0660 <0.0001 ACCDENS 0.0069 0.0048 0.1507 SIGDENS 0.1002 0.0523 0.0556 Dispersion (k) 0.4890 -- -- Comments: The p-value for ACCDENS is higher than desirable. Region is included for NC or MN; a value of 0 if in northern or southern CA. ACCDENS = number of driveways plus unsignalized intersections per mile. SIGDENS = number of signalized intersections per mile. Table E-14. Alternate Model 1 for commercial total crashes. Variable Estimate Standard Error P-Value Intercept -0.6854 0.5010 0.1713 Region 0.6166 0.1013 <0.0001 AADT 0.3766 0.0468 <0.0001 PROPNODEV -0.4252 0.2268 0.0608 Dispersion (k) 0.5165 -- -- Comments: This model was developed from the dataset combining all land use types with factor variables representing the land use. Region is included for NC or MN; a value of 0 if in northern or southern CA. PROPNODEV = proportion of length with no roadside development. Table E-15. Alternate Model 2 for commercial total crashes. Injury Crashes Model form: Crashes/mile/year = exp(intercept+Region) ∗ (AADT)b ∗ exp(c1∗X1+ . . . +cn∗Xn) Variable Estimate Standard Error P-Value Intercept -2.0602 0.7991 0.0099 Region 0.4672 0.1815 0.0100 AADT 0.3649 0.0766 <0.0001 ACCDENS 0.0085 0.0047 0.0679 SIGDENS 0.0566 0.0512 0.2696 Dispersion (k) 0.4406 -- -- Comments: The p-value for SIGDENS is higher than desirable. Region is included for NC or MN; a value of 0 if in northern or southern CA. ACCDENS = number of driveways plus unsignalized intersections per mile. SIGDENS = number of signalized intersections per mile. Table E-16. Alternate Model 1 for commercial injury crashes.

162 Application of Crash Modification Factors for Access Management Variable Estimate Standard Error P-Value Intercept -0.9792 0.8386 0.2430 Region 0.2383 0.1497 0.1113 AADT 0.3225 0.0797 <0.0001 PROPNODEV -0.6472 0.3040 0.0333 PROPLANE1 -0.6047 0.2631 0.0216 Dispersion (k) 0.4228 -- -- Comments: Region is included for NC or MN; a value of 0 if in northern or southern CA. PROPLANE1 = proportion of corridor length with two lanes. PROPNODEV = proportion of length with no roadside development. Table E-17. Alternate Model 2 for commercial injury crashes. Variable Estimate Standard Error P-Value Intercept 0.2127 0.7288 0.7704 Region 0.6769 0.1559 <0.0001 AADT 0.2705 0.0697 0.0001 PROPVC 0.5421 0.1990 0.0064 PROPLANE1 -0.6244 0.2566 0.0150 Dispersion (k) 0.4739 -- -- Comments: Region is included for NC or MN; a value of 0 if in northern or southern CA. PROPLANE1 = proportion of corridor length with two lanes. PROPVC = proportion of length with visual clutter. Table E-18. Alternate Model 3 for commercial injury crashes. Variable Estimate Standard Error P-Value Intercept -1.9690 0.5862 0.0008 Region 0.3056 0.0923 0.0009 AADT 0.3751 0.0548 <0.0001 SIGDENS 0.1075 0.0300 0.0003 PROPLANE1 -0.5245 0.1430 0.0002 Dispersion (k) 0.3951 -- -- Comments: This model was developed from the dataset combining all land use types with factor variables representing the land use. Region is included for NC or MN; a value of 0 if in northern or southern CA. PROPLANE1 = proportion of corridor length with two lanes. SIGDENS = number of signalized intersections per mile. Table E-19. Alternate Model 4 for commercial injury crashes.

Summary of Models by Land Use and Crash Type 163   Turning Crashes Model form: Crashes/mile/year = exp(intercept+Region) ∗ (AADT)b ∗ exp(c1∗X1+ . . . +cn∗Xn) Variable Estimate Standard Error P-Value Intercept -0.9816 0.9366 0.2946 Region 0 -- -- AADT 0.1650 0.0960 0.0855 ACCDENS 0.0110 0.0052 0.0359 SIGDENS 0.1995 0.0660 0.0025 Dispersion (k) 0.7140 -- -- Comments: Region is not included; a value of 0 is assumed for all regions. ACCDENS = number of driveways plus unsignalized intersections per mile. SIGDENS = number of signalized intersections per mile. Table E-20. Alternate Model 1 for commercial turning crashes. Variable Estimate Standard Error P-Value Intercept 0.0085 1.1277 0.9940 Region -0.2548 0.2101 0.2251 AADT 0.1947 0.1068 0.0685 PROPNODEV -0.6967 0.4150 0.0932 PROPLANE1 -0.7328 0.3577 0.0405 Dispersion (k) 0.7802 -- -- Comments: Region is included for NC or MN; a value of 0 if in northern or southern CA. PROPLANE1 = proportion of corridor length with two lanes. PROPNODEV = proportion of length with no roadside development. Table E-21. Alternate Model 2 for commercial turning crashes. Rear-End Crashes Model form: Crashes/mile/year = exp(intercept+Region) ∗ (AADT)b ∗ exp(c1∗X1+ . . . +cn∗Xn) Variable Estimate Standard Error P-Value Intercept -3.2746 0.8502 0.0001 Region 0.8114 0.1786 <0.0001 AADT 0.5050 0.0827 <0.0001 SIGDENS 0.0924 0.0552 0.0941 Dispersion (k) 0.6055 -- -- Comments: Region is included for NC or MN; a value of 0 if in northern or southern CA. SIGDENS = number of signalized intersections per mile. Table E-22. Alternate Model 1 for commercial rear-end crashes.

164 Application of Crash Modification Factors for Access Management Right-Angle Crashes Model form: Crashes/mile/year = exp(intercept+Region) ∗ (AADT)b ∗ exp(c1∗X1+ . . . +cn∗Xn) Variable Estimate Standard Error P-Value Intercept -3.0651 0.6691 <0.0001 Region 0.8113 0.1136 <0.0001 AADT 0.5015 0.0618 <0.0001 PROPLANE1 -0.5548 0.1713 0.0012 SIGDENS 0.0621 0.0380 0.1021 Dispersion (k) 0.6098 -- -- Comments: This model was developed from the dataset combining all land use types with factor variables representing the land use. Region is included for NC or MN; a value of 0 if in northern or southern CA. PROPLANE1 = proportion of corridor length with two lanes. SIGDENS = number of signalized intersections per mile. Table E-23. Alternate Model 2 for commercial rear-end crashes. Variable Estimate Standard Error P-Value Intercept -1.6746 0.9312 0.0721 Region 1.4756 0.2388 <0.0001 AADT 0.1238 0.0912 0.1745 ACCDENS 0.0165 0.0064 0.0099 SIGDENS 0.1532 0.0658 0.0199 Dispersion (k) 0.7288 -- -- Comments: The p-value for AADT is higher than desirable. Region is included for NC or MN; a value of 0 if in northern or southern CA. ACCDENS = number of driveways plus unsignalized intersections per mile. SIGDENS = number of signalized intersections per mile. Table E-24. Alternate Model 1 for commercial right-angle crashes. Variable Estimate Standard Error P-Value Intercept -1.9023 0.6838 0.0054 Region 1.2344 0.1377 <0.0001 AADT 0.2433 0.0648 0.0002 PROPFULLDEV 0.6787 0.1846 0.0002 Dispersion (k) 0.7674 -- -- Comments: This model was developed from the dataset combining all land use types with factor variables representing the land use. Region is included for NC or MN; a value of 0 if in northern or southern CA. PROPFULLDEV = proportion of corridor length with full roadside development. Table E-25. Alternate Model 2 for commercial right-angle crashes.

Summary of Models by Land Use and Crash Type 165   Residential Models Total Crashes Model form: Crashes/mile/year = exp(intercept+Region) ∗ (AADT)b ∗ exp(c1∗X1+ . . . +cn∗Xn) Variable Estimate Standard Error P-Value Intercept -0.5615 0.7076 0.4275 Region 0.4443 0.1533 0.0038 AADT 0.3094 0.0673 <0.0001 PROPLANE1 -0.5479 0.1702 0.0013 SIGDENS 0.1262 0.0629 0.0449 PROPFULLDEV 0.3371 0.2317 0.1456 Dispersion (k) 0.3277 -- -- Comments: The p-value for PROPFULLDEV is larger than desirable. Region is included for NC or MN; a value of 0 if in northern or southern CA. PROPFULLDEV = proportion of corridor length with full roadside development. PROPLANE1 = proportion of corridor length with two lanes. SIGDENS = number of signalized intersections per mile. Table E-26. Alternate Model 1 for residential total crashes. Variable Estimate Standard Error P-Value Intercept -0.4764 0.7211 0.5088 Region 0.3824 0.1499 0.0108 AADT 0.3025 0.0685 <0.0001 PROPLANE1 -0.5260 0.1722 0.0023 SIGDENS 0.1576 0.0622 0.0113 Dispersion (k) 0.3384 -- -- Comments: This model is the same as Alternate Model 1, but without PROPFULLDEV. Region is included for NC or MN; a value of 0 if in northern or southern CA. PROPLANE1 = proportion of corridor length with two lanes. SIGDENS = number of signalized intersections per mile. Table E-27. Alternate Model 2 for residential total crashes. Variable Estimate Standard Error P-Value Intercept -1.3644 0.4953 0.0059 Region 0.6850 0.1107 <0.0001 AADT 0.3883 0.0463 <0.0001 ACCDENS 0.0032 0.0022 0.1375 Dispersion (k) 0.5181 -- -- Comments: This model was developed from the dataset combining all land use types with factor variables representing the land use. Region is included for NC or MN; a value of 0 if in northern or southern CA. ACCDENS = number of driveways plus unsignalized intersections per mile. Table E-28. Alternate Model 3 for residential total crashes.

166 Application of Crash Modification Factors for Access Management Injury Crashes Model form: Crashes/mile/year = exp(intercept+Region) ∗ (AADT)b ∗ exp(c1∗X1+ . . . +cn∗Xn) Variable Estimate Standard Error P-Value Intercept -1.1048 0.4876 0.0235 Region 0.6166 0.1013 <0.0001 AADT 0.3766 0.0468 <0.0001 PROPNODEV -0.4252 0.2268 0.0608 Dispersion (k) 0.5165 -- -- Comments: This model was developed from the dataset combining all land use types with factor variables representing the land use. Region is included for NC or MN; a value of 0 if in northern or southern CA. PROPNODEV = proportion of length with no roadside development. Table E-29. Alternate Model 4 for residential total crashes. Variable Estimate Standard Error P-Value Intercept -2.7357 0.8556 0.0014 Region 0.1656 0.1423 0.2447 AADT 0.4189 0.0820 <0.0001 PROPLANE1 -0.4040 0.1669 0.0155 SIGDENS 0.2081 0.0539 0.0001 Dispersion (k) 0.2663 -- -- Comments: Region is included for NC or MN; a value of 0 if in northern or southern CA. PROPLANE1 = proportion of corridor length with two lanes. SIGDENS = number of signalized intersections per mile. Table E-30. Alternate Model 1 for residential injury crashes. Variable Estimate Standard Error P-Value Intercept -2.7379 0.9147 0.0028 Region 0.2303 0.1603 0.1509 AADT 0.4615 0.0867 <0.0001 PROPLANE1 -0.6125 0.1715 0.0004 PROPFULLDEV 0.3720 0.2273 0.1017 Dispersion (k) 0.3220 -- -- Comments: Region is included for NC or MN; a value of 0 if in northern or southern CA. PROPFULLDEV = proportion of corridor length with full roadside development. PROPLANE1 = proportion of corridor length with two lanes. Table E-31. Alternate Model 2 for residential injury crashes.

Summary of Models by Land Use and Crash Type 167   Turning Crashes Model form: Crashes/mile/year = exp(intercept+Region) ∗ (AADT)b ∗ exp(c1∗X1+ . . . +cn∗Xn) Variable Estimate Standard Error P-Value Intercept -2.5087 1.0439 0.0163 Region 0 -- -- AADT 0.2949 0.1008 0.0034 UNSIGDENS 0.0589 0.0289 0.0416 SIGDENS 0.2173 0.0845 0.0101 Dispersion (k) 0.6710 -- -- Comments: Region is not included; a value of 0 is assumed for all regions. SIGDENS = number of signalized intersections per mile. UNSIGDENS = number of unsignalized intersections per mile. Table E-32. Alternate Model 1 for residential turning crashes. Variable Estimate Standard Error P-Value Intercept -1.1275 1.1225 0.3152 Region -0.6520 0.2073 0.0017 AADT 0.1826 0.1059 0.0846 UNSIGDENS 0.0635 0.0283 0.0247 SIGDENS 0.2244 0.0818 0.0061 Dispersion (k) 0.5792 -- -- Comments: This model is the same as Alternate Model 1, but with the region variable included. Note the large reduction in the AADT parameter. Region is included for NC or MN; a value of 0 if in northern or southern CA. SIGDENS = number of signalized intersections per mile. UNSIGDENS = number of unsignalized intersections per mile. Table E-33. Alternate Model 2 for residential turning crashes. Variable Estimate Standard Error P-Value Intercept -0.9528 0.7286 0.1910 Region -0.1651 0.1339 0.2174 AADT 0.1759 0.0708 0.0130 ACCDENS 0.0052 0.0028 0.0643 SIGDENS 0.1821 0.0426 <0.0001 Dispersion (k) 0.7030 -- -- Comments: This model was developed from the dataset combining all land use types with factor variables representing the land use. Region is included for NC or MN; a value of 0 if in northern or southern CA. ACCDENS = number of driveways plus unsignalized intersections per mile. SIGDENS = number of signalized intersections per mile. Table E-34. Alternate Model 3 for residential turning crashes.

168 Application of Crash Modification Factors for Access Management Rear-End Crashes Model form: Crashes/mile/year = exp(intercept+Region) ∗ (AADT)b ∗ exp(c1∗X1+ . . . +cn∗Xn) Variable Estimate Standard Error P-Value Intercept -0.7154 0.7477 0.3387 Region -0.3163 0.1301 0.0150 AADT 0.2179 0.0729 0.0028 PROPNODEV -0.5890 0.2827 0.0372 Dispersion (k) 0.7791 -- -- Comments: This model was developed from the dataset combining all land use types with factor variables representing the land use. Region is included for NC or MN; a value of 0 if in northern or southern CA. PROPNODEV = proportion of length with no roadside development. Table E-35. Alternate Model 4 for residential turning crashes. Variable Estimate Standard Error P-Value Intercept -3.8941 0.9816 <0.0001 Region 0.5803 0.1984 0.0034 AADT 0.5392 0.0945 <0.0001 SIGDENS 0.1675 0.0864 0.0527 Dispersion (k) 0.5541 -- -- Comments: Region is included for NC or MN; a value of 0 if in northern or southern CA. SIGDENS = number of signalized intersections per mile. Table E-36. Alternate Model 1 for residential rear-end crashes. Variable Estimate Standard Error P-Value Intercept -2.6180 1.0221 0.0104 Region 0.5406 0.1865 0.0037 AADT 0.4782 0.0967 <0.0001 PROPLANE1 -0.8174 0.2078 <0.0001 PROPTWLTL -0.5600 0.2439 0.0217 Dispersion (k) 0.4803 -- -- Comments: Region is included for NC or MN; a value of 0 if in northern or southern CA. PROPLANE1 = proportion of corridor length with two lanes. PROPTWLTL = proportion of corridor length with TWLTL. Table E-37. Alternate Model 2 for residential rear-end crashes.

Summary of Models by Land Use and Crash Type 169   Right-Angle Crashes Model form: Crashes/mile/year = exp(intercept+Region) ∗ (AADT)b ∗ exp(c1∗X1+ . . . +cn∗Xn) Variable Estimate Standard Error P-Value Intercept -3.3056 0.6549 <0.0001 Region 0.8113 0.1136 <0.0001 AADT 0.5015 0.0618 <0.0001 PROPLANE1 -0.5548 0.1713 0.0012 SIGDENS 0.0621 0.0380 0.1021 Dispersion (k) 0.6098 -- -- Comments: This model was developed from the dataset combining all land use types with factor variables representing the land use. Region is included for NC or MN; a value of 0 if in northern or southern CA. PROPLANE1 = proportion of corridor length with two lanes. SIGDENS = number of signalized intersections per mile. Table E-38. Alternate Model 3 for residential rear-end crashes. Variable Estimate Standard Error P-Value Intercept -1.8958 1.1271 0.0926 Region 0.8655 0.2364 0.0003 AADT 0.2357 0.1098 0.0319 Dispersion (k) 0.7812 -- -- Comments: Region is included for NC or MN; a value of 0 if in northern or southern CA. Table E-39. Alternate Model 1 for residential right-angle crashes. Variable Estimate Standard Error P-Value Intercept -1.4079 1.0732 0.1896 Region 0.8858 0.2180 <0.0001 AADT 0.1332 0.1051 0.2049 SIGDENS 0.2267 0.0750 0.0025 PROPLANE1 -0.3633 0.2383 0.1274 PROPFULLDEV 0.4295 0.3125 0.1693 Dispersion (k) 0.5555 -- -- Comments: The p-values for AADT, PROPLANE1, and PROPFULLDEV are higher than desirable. Region is included for NC or MN; a value of 0 if in northern or southern CA. PROPFULLDEV = proportion of corridor length with full roadside development. PROPLANE1 = proportion of corridor length with two lanes. SIGDENS = number of signalized intersections per mile. Table E-40. Alternate Model 2 for residential right-angle crashes.

170 Application of Crash Modification Factors for Access Management Variable Estimate Standard Error P-Value Intercept -2.1173 0.6540 0.0012 Region 1.1970 0.1314 <0.0001 AADT 0.1768 0.0639 0.0057 SIGDENS 0.2084 0.0390 <0.0001 ACCDENS 0.0044 0.0028 0.1078 Dispersion (k) 0.6790 -- -- Comments: This model was developed from the dataset combining all land use types with factor variables representing the land use. Region is included for NC or MN; a value of 0 if in northern or southern CA. ACCDENS = number of driveways plus unsignalized intersections per mile. SIGDENS = number of signalized intersections per mile. Table E-41. Alternate Model 3 for residential right-angle crashes.