Color Effectiveness of Yellow Pavement Marking Materials: Full Report (2008)

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.

123 APPENDIX D: FIELD EXPERIMENT Table 24. GEE Analysis Summary for the Binary Data Obtained in the Field Experiment. Note: Statistically significant factors are shown in bold typeface. Skipline represents pavement marking location. Pairwise Material Comparisons Thermoplastic materials from Ennis Paint are denoted as EnnisThermoY++, EnnisThermoY+, and EnnisThermoY, in decreasing order of yellow saturation. Henceforth in this report, EnnisThermoY++ refers to the most saturated yellow thermoplastic material (with 1% yellow organic pigment and 1.4% Titanium dioxide), EnnisThermoY+ refers to the thermoplastic with intermediate yellow saturation (with 1% yellow organic pigment and 2.3% Titanium dioxide), and finally EnnisPaintY refers to the thermoplastic material with relatively lower yellow saturation (with 1% yellow organic pigment and 2.9% Titanium dioxide). gee(formula = Response ~ Subject + Headlamp * Skipline + Headlamp * Material + Skipline * Material, cluster = Subject, variance = "glm.scale", data = NCHRP518Field, family = binomial, link = "logit") Model: Family: binomial Link : logit Estimated Parameters: Regression Coefficients: Estimate Std.Err. Z Prob (Intercept) -2.0543770 0.4678822 -4.39 0.0000113 Subject 0.0078463 0.0240755 0.33 0.7444962 Headlamp -0.1080516 0.1398585 -0.77 0.4397724 Skipline 0.4298452 0.0629424 6.83 0.0000000 Material1 0.7381654 0.2649007 2.79 0.0053268 Material2 0.4375480 0.0983541 4.45 0.0000086 Material3 0.4246154 0.1301054 3.26 0.0011000 Material4 -0.0409369 0.0421986 -0.97 0.3319963 Headlamp:Skipline 0.0432295 0.0338825 1.28 0.2020038 HeadlampMaterial1 -0.0567028 0.0687353 -0.82 0.4094028 HeadlampMaterial2 -0.0079999 0.0438471 -0.18 0.8552301 HeadlampMaterial3 0.0257604 0.0301984 0.85 0.3936366 HeadlampMaterial4 0.0090844 0.0286120 0.32 0.7508622 SkiplineMaterial1 -0.1626559 0.0668595 -2.43 0.0149825 SkiplineMaterial2 -0.1137373 0.0224581 -5.06 0.0000004 SkiplineMaterial3 -0.0666701 0.0290896 -2.29 0.0219121 SkiplineMaterial4 -0.0132592 0.0148220 -0.89 0.3710217 Scale Parameter: 1.004171 Number of iterations : 2 Number of observations : 1560 Number of clusters : 26

124 Ennis Thermoplastic Y++ vs. Ennis Thermoplastic Y+: Pavement marking location (shown as skipline in Table 25) and material type were both statistically significant (p<0.001 and p=0.005, respectively). The further away the samples, the less yellow they appeared. The interaction between material type and location was also significant. Table 25. GEE Summary Table for the Pairwise Comparison between Ennis ThemoplasticY++ vs. Ennis Thermoplastic Y+. The trend in responses with increasing distance (and decreasing observation angle and increasing entrance angle) was not a surprise. For the samples at 180ft, there were more “white” responses than “yellow” responses. For shorter distances, the more saturated yellow markings elicited more “yellow” responses, especially at a distance of 60ft, the difference in yellow response percentages between the two materials were notable. However, with increasing distance, the discrepancy between the percentages diminishes, and even turns in favor of Ennis Thermoplastic Y+ in such far distances. The differences in the trends between the two materials with respect to distance from observers is statistically significant (p=0.015) as indicated by the interaction between skipline and material in Table 25. The more saturated yellow pavement markings seem to be highly efficient at shorter distances in rendering “yellow”, but with increasing distance, this effectiveness becomes less distinct when compared to its less saturated counterpart. Figure 48 shows a bar plot of yellow response percentages for the two materials as a function of distance (location). Estimate Std.Err. Z Prob (Intercept) -2.8547921 0.5714597 -5.00 0.0000006 Subject 0.0065356 0.0247716 0.26 0.7919086 Headlamp 0.0273087 0.2209851 0.12 0.9016502 Skipline 0.6225757 0.0922764 6.75 0.0000000 Material 0.7410158 0.2645802 2.80 0.0050988 Headlamp:Skipline 0.0013387 0.0561586 0.02 0.9809826 Headlamp:Material -0.0626491 0.0687658 -0.91 0.3622693 Skipline:Material -0.1630064 0.0667931 -2.44 0.0146683

125 32.7% 36.5% 44.2% 76.9% 94.2%94.2% 36.5% 40.4% 48.1% 73.1%71.2% 84.6% 0.0% 10.0% 20.0% 30.0% 40.0% 50.0% 60.0% 70.0% 80.0% 90.0% 100.0% 1 2 3 4 5 6 Skipline No Pe rc en ta ge o f Y el lo w R es po ns es EnnisThermoY++ EnnisThermoY+ N=624 Skipline represents pavement marking location, 1 being the closest to the observer at 30ft and 6 being the farthest at 180ft. Note: Skipline No represents pavement marking location, 1 being the closest to the observers at 30ft and 6 being the farthest at 180ft. Figure 48. Percentage of “yellow” responses for Ennis ThermoplasticY++ vs. Ennis Thermoplastic Y+ as a function of pavement marking location. Ennis Thermoplastic Y++ vs. Ennis Thermoplastic Y: Pavement marking location (shown as skipline in Table 26) and material type were both statistically significant (p<0.001 for both). Results were similar to those given in the previous section for the comparison of Ennis Thermoplastic Y++ and Ennis Thermoplastic Y+. GEE summary table is given in Table 26. Table 26. GEE Summary Table for the Pairwise Comparison between Ennis ThemoplasticY++ vs. Ennis Thermoplastic Y.

126 For the samples at 180ft, there were more “white” responses than “yellow” responses. For shorter distances, the more saturated yellow markings elicited more “yellow” responses, especially at a distance of 60ft, the difference in yellow response percentages between the two materials were notable. The reverse trend in responses as the distance increases was more noteworthy between these two materials. Interestingly enough, unlike shorter distances, the less saturated yellow elicited more yellow responses at far distances. Hence, the interaction between the location and the material type was statistically significant (p<0.001). Figure 49 shows a bar plot of yellow response percentages for the two materials as a function of distance (location). Regression Coefficients: Estimate Std.Err. Z Prob (Intercept) -2.6372694 0.6062992 -4.35 0.0000136 Subject 0.0109395 0.0302159 0.36 0.7173194 Headlamp 0.2006379 0.1940610 1.03 0.3011872 Skipline 0.5353335 0.0904947 5.92 0.0000000 Material 1.0319068 0.2327207 4.43 0.0000092 Headlamp:Skipline -0.0397396 0.0515937 -0.77 0.4411567 Headlamp:Material -0.0545756 0.0709002 -0.77 0.4414463 Skipline:Material -0.2529212 0.0490627 -5.16 0.0000003 Scale Parameter: 0.9944718

127 32.7% 36.5% 44.2% 76.9% 94.2%94.2% 44.2%46.2% 65.4%65.4% 67.3% 76.9% 0% 20% 40% 60% 80% 100% 1 2 3 4 5 6 Skipline No Pe rc en ta ge o f Y el lo w R es po ns es EnnisThermoY++ EnnisThermoY N=624 Skipline represents pavement marking location, 1 being the closest to the observer at 30ft and 6 being the farthest at 180ft. Note: Skipline No represents pavement marking location, 1 being the closest to the observers at 30ft and 6 being the farthest at 180ft. Figure 49. Percentage of “yellow” responses for Ennis ThermoplasticY++ vs. Ennis Thermoplastic Y as a function of pavement marking location. Ennis Thermoplastic Y++ vs. Flint Trading Thermoplastic: Pavement marking location (shown as skipline in Table 27) and material type were both statistically significant (p<0.001 for both). Results were similar to those given in the previous pairwise comparisons. The GEE summary table for the comparison between Ennis Thermoplastic Y++ and Flint Trading thermoplastic materials is given in Table 27. Table 27. GEE Summary Table for the Pairwise Comparison between Ennis ThemoplasticY++ vs. Flint Trading Thermoplastic.

128 The general trend for both materials was to be identified as more “white” as the distance increased. Yet, the situation was more dramatic for Ennis Thermoplastic Y++. Again, similar to the earlier pairwise comparisons, the highly saturated yellow thermoplastic (Ennis Thermoplastic Y++) educed higher yellow responses for distances up to 90ft, but beyond 90ft, the trend was reversed. For the samples at 180ft, there were more “white” responses than “yellow” responses. Hence, the interaction between the location and the material type was statistically significant (p<0.001). Figure 50 shows a bar plot of yellow response percentages for the two materials as a function of distance (location). In general, the two materials behaved differently in terms of color rendition with distance. The less-saturated material (Flint Trading Thermoplastic) was more uniformly identified as yellow as compared to the highly saturated thermoplastic material. Regression Coefficients: Estimate Std.Err. Z Prob (Intercept) -3.0680132 0.6115840 -5.02 0.0000005 Subject 0.0184761 0.0274823 0.67 0.5013981 Headlamp -0.0009058 0.1490214 -0.01 0.9951502 Skipline 0.5831072 0.0860656 6.78 0.0000000 Material 0.6989758 0.2000400 3.49 0.0004755 Headlamp:Skipline 0.0245047 0.0446972 0.55 0.5835279 Headlamp:Material 0.0018457 0.0883248 0.02 0.9833279 Skipline:Material -0.2050196 0.0529928 -3.87 0.0001094 Scale Parameter: 0.9950483

129 32.7% 36.5% 44.2% 76.9% 94.2%94.2% 48.1% 53.8% 63.5% 75.0% 76.9% 86.5% 0% 20% 40% 60% 80% 100% 1 2 3 4 5 6 Skipline No Pe rc en ta ge o f Y el lo w R es po ns es EnnisThermoY++ Flint Trading Thermo N=624 Skipline represents pavement marking location, 1 being the closest to the observer at 30ft and 6 being the farthest at 180ft. Note: Skipline No represents pavement marking location, 1 being the closest to the observers at 30ft and 6 being the farthest at 180ft. Figure 50. Percentage of “yellow” responses for Ennis ThermoplasticY++ vs. Flint Trading Thermoplastic as a function of pavement marking location. Ennis Thermoplastic Y++ vs. Latex Paint: Pavement marking location (shown as skipline in Table 28) and material type were both statistically significant (p<0.001 for both). The GEE summary table for the comparison between Ennis Thermoplastic Y++ and Latex Paint materials is given in Table 28. Table 28. GEE Summary Table for the Pairwise Comparison between Ennis ThemoplasticY++ vs. Latex Paint.

130 The second order interaction between material and location was also statistically significant similar to the previous cases (p<0.001). Yet, unlike the previous cases, the third order interaction between material, location, and headlamp was also statistically significant (p≅0.04). That is, the trends for the interaction between location and material were also headlamp dependent. The general trend for both materials was to be identified as more “white” as the distance increased. Overall, the highly saturated yellow thermoplastic (Ennis Thermoplastic Y++) educed higher yellow responses at all distances but at 5th location (150ft). Nonetheless, the difference in the number of yellow responses was not as pronounced at longer distances as it was for shorter distances. Figure 51 shows a bar plot of yellow response percentages for the two materials as a function of distance (location). Figure 52 shows the percentage of yellow responses partitioned into headlamp types. The significant third order interaction between location, material and headlamp can be seen in Figure 52. Note the responses for each material type at different distances for the two headlamps: For the thermoplastic material (Ennis Thermo Y++), HID illumination elicited more yellow responses than did the TH headlamps in general regardless of the distance. Yet, for the latex paint material, HID headlamp illumination causes fewer yellow responses in shorter distances, whereas for longer distances the trend was reversed. HID headlamps elicited more yellow responses in longer distances regardless of the headlamp type. Estimate Std.Err. Z Prob (Intercept) -1.9430579 0.4721223 -4.12 0.0000386 Subject -0.0107647 0.0253230 -0.43 0.6707662 Headlamp -0.1725517 0.1729792 -1.00 0.3185079 Skipline 0.5206151 0.0698420 7.45 0.0000000 Material 1.4262039 0.3130920 4.56 0.0000052 Headlamp:Skipline 0.0821678 0.0467265 1.76 0.0786653 Headlamp:Material -0.3450977 0.1936340 -1.78 0.0747140 Skipline:Material -0.2663101 0.0747388 -3.56 0.0003663 Headlamp:Skipline:Material 0.1013593 0.0490711 2.07 0.0388702

131 32.7% 36.5% 44.2% 76.9% 94.2%94.2% 25.0% 42.3% 32.7% 55.8% 67.3% 46.2% 0% 20% 40% 60% 80% 100% 1 2 3 4 5 6 Skipline No Pe rc en ta ge o f Y el lo w R es po ns es EnnisThermoY++ Latex Paint N=624 Skipline represents pavement marking location, 1 being the closest to the observer at 30ft and 6 being the farthest at 180ft. Note: Skipline No represents pavement marking location, 1 being the closest to the observers at 30ft and 6 being the farthest at 180ft. Figure 51. Percentage of “yellow” responses for Ennis ThermoplasticY++ vs. Latex Paint as a function of pavement marking location.

132 96 .2 % 76 .9 % 50 .0 % 30 .8 % 3 8. 5% 61 .5 % 53 .8 % 34 .6 % 53 .8 % 38 .5 % 92 .3 % 92 .3 % 76 .9 % 38 .5 % 42 .3 % 26 .9 % 50 .0 % 73 .1 % 57 .7 % 30 .8 % 30 .8 % 11 .5 % 96 .2 % 42 .3 % 0% 20% 40% 60% 80% 100% 1 2 3 4 5 6 Skipline No Pe rc en ta ge o f Y el lo w R es po ns es HID EnnisThermoY++ HID Latex Paint TH EnnisThermoY++ TH Latex Paint N=624 Skipline represents pavement marking location, 1 being the closest to the observer at 30ft and 6 being the farthest at 180ft. Figure 52. Percentage of “yellow” responses for Ennis ThermoplasticY++ vs. Latex Paint as a function of pavement marking location and headlamp type. Ennis Thermoplastic Y+ vs. Ennis Thermoplastic Y: For these two materials, the only statistically significant factor was pavement marking location (p<0.001). Statistically speaking, these two materials performed equally well. The GEE summary table for the comparison between Ennis Thermoplastic Y+ and Ennis Thermoplastic Y materials is given in Table 29. Material type and the interaction between the material type and location was just short of having statistical significance at α=0.05 confidence level (p=0.067 and p=0.052, respectively). Although there was a slight trend similar to the earlier pairwise comparisons for material type and location, the difference was short of having a statistical significance. No other main factor or interaction was found to be statistically significant. Figure 53 shows the percentage of yellow responses for these two material types as a function of pavement marking location.

133 Table 29. GEE Summary Table for the Pairwise Comparison between Ennis Themoplastic Y+ vs. Ennis Thermoplastic Y. Regression Coefficients: Estimate Std.Err. Z Prob (Intercept) -1.9324491 0.5390669 -3.58 0.0003373 Subject 0.0138516 0.0267355 0.52 0.6043913 Headlamp 0.0519879 0.1720962 0.30 0.7625862 Skipline 0.3718200 0.0877815 4.24 0.0000228 Material 0.2847902 0.1554002 1.83 0.0668584 Headlamp:Skipline -0.0183845 0.0393280 -0.47 0.6401658 Headlamp:Material 0.1360718 0.1571672 0.87 0.3866120 Skipline:Material -0.0887882 0.0455997 -1.95 0.0515199 Headlamp:Skipline:Material -0.0327811 0.0411028 -0.80 0.4251376

134 36.5% 40.4% 48.1% 73.1%71.2% 84.6% 44.2%46.2% 65.4%65.4% 67.3% 76.9% 0% 20% 40% 60% 80% 100% 1 2 3 4 5 6 Skipline No Pe rc en ta ge o f Y el lo w R es po ns es EnnisThermoY+ EnnisThermoY N=624 Skipline represents pavement marking location, 1 being the closest to the observer at 30ft and 6 being the farthest at 180ft. Note: Skipline No represents pavement marking location, 1 being the closest to the observers at 30ft and 6 being the farthest at 180ft. Figure 53. Percentage of “yellow” responses for Ennis ThermoplasticY+ vs. Ennis Thermoplastic Y as a function of pavement marking location. Ennis Thermoplastic Y+ vs. Flint Trading Thermoplastic: For these two materials, the only statistically significant factor was again pavement marking location (p<0.001). These two materials were very similar in nature. The GEE summary table for the comparison between Ennis Thermoplastic Y+ and Ennis Thermoplastic Y materials is given in Table 30. No other main factor or interaction was found to be statistically significant. Figure 54 shows the percentage of yellow responses for these two material types as a function of pavement marking location. Table 30. GEE Summary Table for the Pairwise Comparison between Ennis ThemoplasticY+ vs. Flint Trading Thermoplastic.

135 36.5% 40.4% 48.1% 73.1%71.2% 84.6% 48.1% 53.8% 63.5% 75.0% 76.9% 86.5% 0% 20% 40% 60% 80% 100% 1 2 3 4 5 6 Skipline No Pe rc en ta ge o f Y el lo w R es po ns es EnnisThermoY+ Flint Trading Thermo N=624 Skipline represents pavement marking location, 1 being the closest to the observer at 30ft and 6 being the farthest at 180ft. Figure 54. Percentage of “yellow” responses for Ennis Thermoplastic Y+ vs. Flint Trading Thermoplastic material as a function of pavement marking location. Ennis Thermoplastic Y+ vs. Latex Paint: Pavement marking location (shown as skipline in Table 31) and material type were both statistically significant (p<0.001 and p=0.021, respectively). The GEE summary table for the comparison between Ennis Thermoplastic Y+ and Latex Paint materials is given in Table 31. Regression Coefficients: Estimate Std.Err. Z Prob (Intercept) -2.3599131 0.5120934 -4.61 0.0000041 Subject 0.0207929 0.0243483 0.85 0.3931191 Headlamp -0.0967779 0.1692611 -0.57 0.5674799 Skipline 0.4202872 0.0754286 5.57 0.0000000 Material -0.0428596 0.1897291 -0.23 0.8212802 Headlamp:Skipline 0.0337560 0.0446545 0.76 0.4496869 Headlamp:Material -0.0124906 0.1751556 -0.07 0.9431500 Skipline:Material -0.0417148 0.0599002 -0.70 0.4861758 Headlamp:Skipline:Material 0.0193270 0.0403596 0.48 0.6320323

136 Table 31. GEE Summary Table for the Pairwise Comparison between Ennis Themoplastic Y+ vs. Latex Paint. The second order interaction between headlamp and pavement marking location was also statistically significant (p≅0.01). Furthermore, the third order interaction between material, location, and headlamp was also statistically significant (p≅0.04). That is, the trends for the interaction between location and material were also headlamp dependent. The interaction between material and pavement marking location was not statistically significant. Figure 55 illustrates the percentage of yellow responses as a function of location and material type. The interaction between headlamp type and location is apparent in Figure 56. HID headlamp illumination elicited more “white” responses for distances up to 90ft, beyond which the trend was the contrary. Such shift in the response trend may be attributed to higher illumination at longer distances provided by the HID headlamps rather than solely on spectral content. However, as yet, the reason is unknown. Figure 57 shows the percentage of yellow responses categorized into headlamp types. The significant third order interaction between location, material and headlamp can be seen in Figure 57. Note the responses for each headlamp type at different distances for the two materials: under HID headlamp illumination, more subjects voted for yellow for Ennis Thermoplastic Y+ type material only for closer distances. For distances beyond 60ft, the “yellow” responses were similar for both materials under HID illumination. Yet, under TH headlamp illumination, Ennis Thermoplastic Y+ type material had more “yellow” votes that did the Latex paint regardless of the distance. In a sense, HID headlamps pronounce the yellow in latex paint more successfully than did the TH headlamps at longer distances. Regression Coefficients: Estimate Std.Err. Z Prob (Intercept) -1.2657229 0.4213129 -3.00 0.0026625 Subject -0.0057865 0.0234129 -0.25 0.8047914 Headlamp -0.3005275 0.1801009 -1.67 0.0951845 Skipline 0.3568612 0.0704066 5.07 0.0000004 Material 0.6823739 0.2964756 2.30 0.0213567 Headlamp:Skipline 0.0988561 0.0386098 2.56 0.0104556 Headlamp:Material -0.2165778 0.1512392 -1.43 0.1521377 Skipline:Material -0.1028277 0.0671888 -1.53 0.1259105 Headlamp:Skipline:Material 0.0844810 0.0402295 2.10 0.0357311

137 36.5% 40.4% 48.1% 73.1%71.2% 84.6% 25.0% 42.3% 32.7% 55.8% 67.3% 46.2% 0% 20% 40% 60% 80% 100% 1 2 3 4 5 6 Skipline No Pe rc en ta ge o f Y el lo w R es po ns es EnnisThermoY+ Latex Paint N=624 Skipline represents pavement marking location, 1 being the closest to the observer at 30ft and 6 being the farthest at 180ft. Note: Skipline No represents pavement marking location, 1 being the closest to the observers at 30ft and 6 being the farthest at 180ft. Figure 55. Percentage of “yellow” responses for Ennis ThermoplasticY+ vs. Latex Paint as a function of pavement marking location.

138 67.3% 65.4% 55.8% 42.3% 46.2% 40.4% 63.5% 73.1% 73.1% 38.5% 36.5% 21.2% 0% 20% 40% 60% 80% 100% 1 2 3 4 5 6 Skipline No Pe rc en ta ge o f Y el lo w R es po ns es HID TH N=624 Skipline represents pavement marking location, 1 being the closest to the observer at 30ft and 6 being the farthest at 180ft. Figure 56. Percentage of combined “yellow” responses under HID and TH headlamp illumination for the combined data of the two materials Ennis Thermoplastic Y+ and Latex Paint as a function of pavement marking location.

139 69 .2 % 57 .7 % 50 .0 % 38 .5 % 42 .3 % 61 .5 % 53 .8 % 34 .6 % 53 .8 % 38 .5 % 76 .9 % 73 .1 % 88 .5 % 46 .2 % 42 .3 % 30 .8 % 50 .0 % 73 .1 % 57 .7 % 30 .8 % 30 .8 % 11 .5 % 92 .3 % 42 .3 % 0% 20% 40% 60% 80% 100% 1 2 3 4 5 6 Skipline No Pe rc en ta ge o f Y el lo w R es po ns es HID EnnisThermoY+ HID Latex Paint TH EnnisThermoY+ TH Latex Paint N=624 Skipline represents pavement marking location, 1 being the closest to the observer at 30ft and 6 being the farthest at 180ft. Figure 57. Percentage of “yellow” responses for Ennis ThermoplasticY+ vs. Latex Paint as a function of pavement marking location and headlamp type.

140 Ennis Thermoplastic Y vs. Flint Trading Thermoplastic: For these two materials, both location and material type were statistically significant factors (p<0.001 and p≅0.007, respectively). The GEE summary table for the comparison between Ennis Thermoplastic Y and Flint Trading Thermoplastic materials is given in Table 32. No interactions were found to be statistically significant. Figure 58 shows the percentage of yellow responses for these two material types as a function of pavement marking location. Table 32. GEE Summary Table for the Pairwise Comparison between Ennis Themoplastic Y vs. Flint Trading Thermoplastic Regression Coefficients: Estimate Std.Err. Z Prob (Intercept) -2.1270022 0.5434218 -3.91 0.0000907 Subject 0.0244241 0.0291967 0.84 0.4028527 Headlamp 0.0397262 0.1907855 0.21 0.8350536 Skipline 0.3319076 0.0729135 4.55 0.0000053 Material -0.3289563 0.1229233 -2.68 0.0074482 Headlamp:Skipline 0.0008894 0.0437665 0.02 0.9837870 Headlamp:Material -0.1492236 0.1349175 -1.11 0.2687112 Skipline:Material 0.0473965 0.0402237 1.18 0.2386688 Headlamp:Skipline:Material 0.0523082 0.0358740 1.46 0.1448098

141 44.2%46.2% 65.4%65.4%67.3% 45.5% 48.1% 53.8% 63.5% 75.0% 76.9% 86.5% 0% 20% 40% 60% 80% 100% 1 2 3 4 5 6 Skipline No Pe rc en ta ge o f Y el lo w R es po ns es EnnisThermo Y Flint Trading Thermo N=624 Skipline represents pavement marking location, 1 being the closest to the observer at 30ft and 6 being the farthest at 180ft. Note: Skipline No represents pavement marking location, 1 being the closest to the observers at 30ft and 6 being the farthest at 180ft. Figure 58. Percentage of “yellow” responses for Ennis Thermoplastic Y vs. Flint Trading Thermoplastic material as a function of pavement marking location. Ennis Thermoplastic Y vs. Latex Paint: Among the main factors, only the pavement marking location was statistically significant (p<0.001). Among the second order interactions, only the interaction between headlamp and material type was significant (p≅0.019). Furthermore, the third order interaction between material, headlamp, and pavement marking location was also statistically significant (p<0.001). The GEE summary table for the comparison between Ennis Thermoplastic Y and Latex paint type materials is given in Table 33. Table 33. GEE Summary Table for the Pairwise Comparison between Ennis Themoplastic Y vs. Latex Paint.

142 Figure 58 shows the percentage of yellow responses for these two material types as a function of pavement marking location. 44.2%46.2% 65.4%65.4%67.3% 45.5% 25.0% 42.3% 32.7% 55.8% 67.3% 46.2% 0% 20% 40% 60% 80% 1 2 3 4 5 6 Skipline No Pe rc en ta ge o f Y el lo w R es po ns es EnnisThermo Y Latex Paint N=624 Skipline represents pavement marking location, 1 being the closest to the observer at 30ft and 6 being the farthest at 180ft. Note: Skipline No represents pavement marking location, 1 being the closest to the observers at 30ft and 6 being the farthest at 180ft. Figure 59. Percentage of “yellow” responses for Ennis Thermoplastic Y vs. Latex Paint type material as a function of pavement marking location. HID headlamp illumination yielded more “yellow” responses only for the latex paint type pavement marking. This interaction can be seen in Figure 60. Regression Coefficients: Estimate Std.Err. Z Prob (Intercept) -1.0372759 0.4579796 -2.26 0.0235191 Subject -0.0016194 0.0262698 -0.06 0.9508462 Headlamp -0.1646294 0.1662279 -0.99 0.3219866 Skipline 0.2681352 0.0624103 4.30 0.0000174 Material 0.3979365 0.2545302 1.56 0.1179549 Headlamp:Skipline 0.0661094 0.0391573 1.69 0.0913529 Headlamp:Material -0.3522719 0.1495834 -2.36 0.0185217 Skipline:Material -0.0142034 0.0575952 -0.25 0.8052117 Headlamp:Skipline:Material 0.1171564 0.0336624 3.48 0.0005008

143 61% 47% 61% 42% 0% 25% 50% 75% 100% Ennis Thermo Y Latex Paint HID TH Figure 60. The percentage of yellow responses for the two headlamps as a function of pavement marking type. The trends for the interaction between headlamp and material were location-dependent. The higher number of “yellow” responses under HID illumination was especially pronounced at far distances similar to the cases in earlier pairwise comparisons. Figure 61 shows the percentage of yellow responses categorized into headlamp types. The significant third order interaction between location, material and headlamp can be seen in Figure 61. Note the responses for each headlamp type at different distances for the two materials: there was no significant difference in the number of yellow responses between HID and TH headlamps for Ennis Thermoplastic Y type material regardless of the distance. However, the same was not true for Latex Paint type material. The further away the material, the more yellow responses under HID illumination only. The yellow responses suffered under TH headlamp illumination for distances beyond 60ft.

144 69 .2 % 65 .4 % 61 .5 % 46 .2 % 42 .3 % 61 .5 % 53 .8 % 34 .6 % 53 .8 % 38 .5 % 73 .1 % 65 .4 % 65 .4 % 69 .2 % 46 .2 % 46 .2 % 50 .0 % 73 .1 % 57 .7 % 30 .8 % 30 .8 % 11 .5 % 80 .8 % 42 .3 % 0% 20% 40% 60% 80% 100% 1 2 3 4 5 6 Skipline No Pe rc en ta ge o f Y el lo w R es po ns es HID EnnisThermo Y HID Latex Paint TH EnnisThermo Y TH Latex Paint N=624 Skipline represents pavement marking location, 1 being the closest to the observer at 30ft and 6 being the farthest at 180ft. Figure 61. Percentage of “yellow” responses for Ennis Thermoplastic Y vs. Latex Paint as a function of pavement marking location and headlamp type. Flint Trading Thermoplastic vs. Latex Paint: Among the main factors, both pavement marking location and material type were statistically significant (p<0.001 and p≅0.002, respectively). Among the second order interactions, only the interaction between headlamp and pavement marking location was statistically significant (p≅0.015). Furthermore, the third order interaction between material, headlamp, and pavement marking location was also statistically significant (p<0.001). The GEE summary table for the comparison between Ennis Thermoplastic Y and Latex paint type materials is given in Table 34. Table 34. GEE Summary Table for the Pairwise Comparison between Flint Trading Thermoplastic vs. Latex Paint.

145 Figure 62 shows the percentage of yellow responses for these two material types as a function of pavement marking location. 48.1% 53.8% 63.5% 75.0%76.9% 86.5% 25.0% 42.3% 32.7% 55.8% 67.3% 46.2% 0% 20% 40% 60% 80% 100% 1 2 3 4 5 6 Skipline No Pe rc en ta ge o f Y el lo w R es po ns es Flint Trading Thermo Latex Paint N=624 Skipline represents pavement marking location, 1 being the closest to the observer at 30ft and 6 being the farthest at 180ft. Note: Skipline No represents pavement marking location, 1 being the closest to the observers at 30ft and 6 being the farthest at 180ft. Figure 62. Percentage of “yellow” responses for Flint Trading Thermoplastic vs. Latex Paint type material as a function of pavement marking location. The second order interaction between headlamp and material type was just short of having statistical significance at α=0.05 significance level. Estimate Std.Err. Z Prob (Intercept) -1.4414560 0.4167374 -3.46 0.0005424 Subject 0.0040774 0.0236297 0.17 0.8630036 Headlamp -0.3128407 0.2035606 -1.54 0.1243317 Skipline 0.3153469 0.0585939 5.38 0.0000001 Material -0.7250900 0.2357943 -3.08 0.0021044 Headlamp:Skipline 0.1180451 0.0485663 2.43 0.0150740 Headlamp:Material 0.2041531 0.1165149 1.75 0.0797457 Skipline:Material 0.0613690 0.0570298 1.08 0.2818882 Headlamp:Skipline:Material -0.0652531 0.0262752 -2.48 0.0130117

146 As the second order interaction between the headlamp type and pavement marking location suggests, there was a discrepancy between the percentages of yellow responses from one headlamp to the other as the pavement marking location changed. Although there was not much of a difference in yellow responses for closer distances, for pavement markings beyond 90ft, HID headlamps elicited more yellow responses. This second order interaction can be seen in Figure 63. 67.3% 69.2% 59.6% 50.0% 55.8% 48.1% 65.4% 75.0% 71.2% 46.2% 40.4% 25.0% 0% 20% 40% 60% 80% 100% 1 2 3 4 5 6 Skipline No Pe rc en ta ge o f Y el lo w R es po ns es HID TH N=624 Skipline represents pavement marking location, 1 being the closest to the observer at 30ft and 6 being the farthest at 180ft. Figure 63. The percentage of yellow responses for the two headlamps as a function of pavement marking location for the combined data of the two materials Flint Trading Thermoplastic and Latex Paint.