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.
APPENDIX H STATISTICAL TESTING OF APPROACH-LEVEL SAFETY MODELS Entering/Circulating Collisions Model 1 Dependent Variable entercirc Observations Used 100 Standard Wald 95% Confidence Chi- Parameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -13.2495 2.8978 -18.9291 -7.5699 20.91 <.0001 logaadt 1 1.0585 0.3096 0.4517 1.6652 11.69 0.0006 logcircaadt 1 0.3672 0.2283 -0.0803 0.8146 2.59 0.1078 Dispersion 1 1.6650 0.6546 0.7704 3.5982 Model 2 Dependent Variable entercirc Observations Used 97 Standard Wald 95% Confidence Chi- Parameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -13.0434 3.1036 -19.1263 -6.9605 17.66 <.0001 logaadt 1 0.9771 0.3342 0.3221 1.6321 8.55 0.0035 logcircaadt 1 0.3088 0.2469 -0.1752 0.7928 1.56 0.2112 Entry_radius 1 0.0099 0.0044 0.0013 0.0185 5.09 0.0241 Dispersion 1 1.6635 0.6865 0.7409 3.7350 Notes: 1) Dispersion is same as AADT only model. 2) Larger entry radii associated with more accidents. Model 3 Dependent Variable entercirc Observations Used 100 Standard Wald 95% Confidence Chi- Parameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -12.2601 2.8528 -17.8514 -6.6687 18.47 <.0001 logaadt 1 0.9217 0.3117 0.3108 1.5327 8.74 0.0031 logcircaadt 1 0.2900 0.2268 -0.1545 0.7345 1.63 0.2010 Central_island_diame 1 -0.0076 0.0056 -0.0187 0.0034 1.84 0.1749 Entry_width 1 0.0582 0.0348 -0.0100 0.1263 2.80 0.0942 Dispersion 1 1.4949 0.6077 0.6738 3.3164 Notes: 1) Dispersion is lower than AADT only model. 2) Larger entry radii associated with more accidents, t-stat 1.67. 3) Larger central island diameter associated with fewer accidents, t-stat 1.36. NCHRP Web-Only Document 94: Appendixes to NCHRP Report 572: Roundabouts in the United States H-1
Model 4 Dependent Variable entercirc Observations Used 97 Standard Wald 95% Confidence Chi- Parameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -13.0579 3.0605 -19.0564 -7.0595 18.20 <.0001 logaadt 1 1.0048 0.3306 0.3568 1.6528 9.24 0.0024 logcircaadt 1 0.3142 0.2414 -0.1590 0.7874 1.69 0.1932 Entry_radius 1 0.0103 0.0043 0.0018 0.0188 5.64 0.0175 Central_island_diame 1 -0.0046 0.0053 -0.0149 0.0057 0.78 0.3782 Dispersion 1 1.5136 0.6591 0.6447 3.5536 Notes: 1) Dispersion is similar to central island diameter and entry radius model. 2) Larger entry radii associated with more accidents, t-stat 2.40. 3) Larger central island diameter associated with fewer accidents, t-stat 0.87. Model 5 Dependent Variable entercirc Observations Used 93 Standard Wald 95% Confidence Chi- Parameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -8.7613 3.3624 -15.3515 -2.1710 6.79 0.0092 logaadt 1 0.9499 0.3374 0.2886 1.6113 7.93 0.0049 logcircaadt 1 0.2687 0.2516 -0.2243 0.7618 1.14 0.2854 Angle_to_next_leg 1 -0.0425 0.0171 -0.0760 -0.0089 6.16 0.0131 Entry_radius 1 0.0105 0.0042 0.0022 0.0188 6.16 0.0131 Dispersion 1 1.3016 0.6807 0.4670 3.6280 Model 6 Dependent Variable entercirc Observations Used 96 Standard Wald 95% Confidence Chi- Parameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -7.2158 3.0908 -13.2736 -1.1579 5.45 0.0196 logaadt 1 0.7018 0.3000 0.1138 1.2898 5.47 0.0193 logcircaadt 1 0.1321 0.2246 -0.3082 0.5724 0.35 0.5564 Angle_to_next_leg 1 -0.0276 0.0120 -0.0512 -0.0040 5.24 0.0220 Entry_width 1 0.0511 0.0263 -0.0004 0.1027 3.78 0.0518 Dispersion 1 1.0797 0.6333 0.3420 3.4085 NOTE: The negative binomial dispersion parameter was estimated by maximum likelihood. Dependent Variable entercirc Observations Used 86 Analysis Of Parameter Estimates Standard Wald 95% Confidence Chi- Parameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -8.9686 3.4834 -15.7959 -2.1412 6.63 0.0100 logaadt 1 0.8322 0.3639 0.1190 1.5455 5.23 0.0222 logcircaadt 1 0.1370 0.2692 -0.3906 0.6646 0.26 0.6108 var1 1 -138.096 67.1671 -269.741 -6.4505 4.23 0.0398 Dispersion 1 2.0324 0.9133 0.8423 4.9037 Var1=1/entry path radius NCHRP Web-Only Document 94: Appendixes to NCHRP Report 572: Roundabouts in the United States H-2
Exiting/Circulating Collisions Notes: 1) Exiting and circulating AADT terms are in right direction but are never significant. 2) Larger circulating widths, inscribed diameter and central island diameters associated with more accidents. 2 models were done, one with circulating width and inscribed diameter, the other with circulating width and central island diameter. Model 1 Dependent Variable circexit Observations Used 100 Standard Wald 95% Confidence Chi- Parameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -7.7145 3.8748 -15.3090 -0.1199 3.96 0.0465 logexit 1 0.3413 0.2591 -0.1664 0.8491 1.74 0.1876 logcircaadt 1 0.5172 0.3777 -0.2231 1.2575 1.88 0.1709 Dispersion 1 6.1315 1.5910 3.6872 10.1962 Model 2 Dependent Variable circexit Observations Used 100 Standard Wald 95% Confidence Chi- Parameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -11.6805 3.6262 -18.7879 -4.5732 10.38 0.0013 logexit 1 0.2801 0.2929 -0.2939 0.8541 0.91 0.3389 logcircaadt 1 0.2530 0.3247 -0.3834 0.8894 0.61 0.4359 Inscribed_circle_dia 1 0.0222 0.0071 0.0083 0.0361 9.81 0.0017 Circulating_width 1 0.1107 0.0470 0.0185 0.2028 5.54 0.0186 Dispersion 1 2.7694 0.8379 1.5305 5.0110 Model 3 Dependent Variable circexit Observations Used 100 Standard Wald 95% Confidence Chi- Parameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -11.2447 3.7745 -18.6426 -3.8469 8.88 0.0029 logexit 1 0.3227 0.3085 -0.2820 0.9275 1.09 0.2956 logcircaadt 1 0.3242 0.3393 -0.3409 0.9893 0.91 0.3394 Central_island_diame 1 0.0137 0.0066 0.0006 0.0267 4.23 0.0397 Circulating_width 1 0.1458 0.0510 0.0458 0.2459 8.17 0.0043 Dispersion 1 3.0519 0.9588 1.6488 5.6490 Dependent Variable circexit Observations Used 87 Analysis Of Parameter Estimates Standard Wald 95% Confidence Chi- Parameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -3.8095 4.1461 -11.9357 4.3167 0.84 0.3582 logexit 1 0.2413 0.2889 -0.3249 0.8076 0.70 0.4035 logcircaadt 1 0.5626 0.3788 -0.1798 1.3051 2.21 0.1375 var2 1 372.8710 85.0465 206.1830 539.5590 19.22 <.0001 Dispersion 1 3.3171 0.9706 1.8694 5.8860 Var2=1/circulating path radius NCHRP Web-Only Document 94: Appendixes to NCHRP Report 572: Roundabouts in the United States H-3
Dependent Variable circexit Observations Used 87 Analysis Of Parameter Estimates Standard Wald 95% Confidence Chi- Parameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -9.8334 4.3100 -18.2808 -1.3860 5.21 0.0225 logexit 1 0.6005 0.3025 0.0076 1.1934 3.94 0.0471 logcircaadt 1 0.7471 0.3860 -0.0093 1.5036 3.75 0.0529 var3 1 -387.729 121.1334 -625.146 -150.311 10.25 0.0014 Dispersion 1 4.4295 1.2473 2.5507 7.6922 Var3=1/exit path radius Approaching Collisions (includes rear-ends and loss of control on the approach) 1) Adding approach half width improves the fit marginally. 2) No other variables worked. Model 1 Dependent Variable APP Observations Used 139 Standard Wald 95% Confidence Chi- Parameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -5.6561 1.2119 -8.0313 -3.2808 21.78 <.0001 logaadt 1 0.6036 0.1463 0.3168 0.8903 17.02 <.0001 Dispersion 1 1.3297 0.2870 0.8710 2.0300 Model 2 Dependent Variable APP Observations Used 132 Standard Wald 95% Confidence Chi- Parameter DF Estimate Error Limits Square Pr > ChiSq Intercept 1 -5.1527 1.2429 -7.5886 -2.7167 17.19 <.0001 logaadt 1 0.4613 0.1598 0.1481 0.7744 8.33 0.0039 Approach_half_width 1 0.0301 0.0136 0.0035 0.0567 4.93 0.0265 Dispersion 1 1.2895 0.3022 0.8146 2.0414 NCHRP Web-Only Document 94: Appendixes to NCHRP Report 572: Roundabouts in the United States H-4
