National Academies Press: OpenBook

Multimodal Level of Service Analysis for Urban Streets (2008)

Chapter: Chapter 5 - Auto LOS Model

« Previous: Chapter 4 - Data Collection
Page 62
Suggested Citation:"Chapter 5 - Auto LOS Model." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 62
Page 63
Suggested Citation:"Chapter 5 - Auto LOS Model." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 63
Page 64
Suggested Citation:"Chapter 5 - Auto LOS Model." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 64
Page 65
Suggested Citation:"Chapter 5 - Auto LOS Model." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 65
Page 66
Suggested Citation:"Chapter 5 - Auto LOS Model." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 66
Page 67
Suggested Citation:"Chapter 5 - Auto LOS Model." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 67
Page 68
Suggested Citation:"Chapter 5 - Auto LOS Model." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 68
Page 69
Suggested Citation:"Chapter 5 - Auto LOS Model." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 69
Page 70
Suggested Citation:"Chapter 5 - Auto LOS Model." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 70
Page 71
Suggested Citation:"Chapter 5 - Auto LOS Model." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 71

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.

62 5.1 Model Development Identification of Key Variables A correlation analysis was performed to determine what relationships may exist between the dependent variable (i.e., individual participant ratings of LOS) and a dataset of 78 independent variables represented in the video clips or transformations of said variables (i.e., log of mean travel speed). The correlation analysis revealed that no less than 69 variables had a statistically significant relationship with individual participant ratings of LOS. Exhibit 65 summarizes the correlation analysis, including Kendall’s tau rank correlation coefficients. Some variables have not been included in order to reduce the size of the table. For example, transformations of variables have not been in- cluded because they tend to have the same or similar tau rank correlation coefficient patterns and significance values. Care was taken in selecting explanatory variables included in the modeling effort so as to avoid including variables that were highly correlated with each other. Exhibit 66 illustrates the correlation analysis done of the explanatory variables to understand the relationships be- tween these variables. In this table, the tau rank correlation coefficients are shown for correlations between space mean speed and the previously listed explanatory variables. In this table, all significance values are <0.10, meaning the relation- ships are statistically significant at the 90% level. Linear Regression Tests Linear regression techniques were first explored to deter- mine if a multiple linear relationship might exist that could estimate the mean rating obtained for each video clip shown in Phase I and/or II of the study. Independent variables to be used to estimate the depend- ent variable (mean clip rating) were selected from the larger set of explanatory variables. This was done by controlling for redundant explanatory variables (e.g., average travel speed and number of stops) and by retaining those explanatory variables that were highly correlated with the mean clip rat- ing. The explanatory variables included in the stepwise regression exercise were as follows: • Space mean speed, • Number of stops, • Stops per mile, • Presence of median, • Presence of exclusive left-turn lane, • Presence of trees rating, and • Pavement quality rating. Forward stepwise regression techniques were used to allow for the inclusion of variables into the model only if they could increase the ability of the model to predict the dependent variable shown through the increase in R-square value. The results of the stepwise multiple linear regression are shown in Exhibit 67. The adjusted R-square value for the overall model is 0.673. The model is Mean Auto LOS = 3.8 − 0.530(Stops) − 0.155(Median) + 0.355(Left-Turn Lane) + 0.098(Trees) + 0.205(Pavement Quality) (Eq. 9) Where Mean Auto LOS = 6.0 for LOS A and 1.0 for LOS F Stops = number of times in video clip that auto speed drops below 5 mph. Median = 3 if raised median (curbs between opposing traffic streams), 2 if two-way left-turn lane, 1 if no opposing traffic stream (one-way street), 0 if no separa- tion between opposing traffic streams. Left-Turn Lane = one if present, zero otherwise. Exclusive left-turn lane can be of any length or C H A P T E R 5 Auto LOS Model

63 Variable tau Rank Correlation Coefficient Significance p-value Space Mean Speed 0.317 0.000 Total Travel Time -0.315 0.000 Lane Width 0.307 0.000 Number of Stops per Mile -0.307 0.000 Sign Quality 0.268 0.000 Tree Presence 0.248 0.000 Sidewalk Width -0.244 0.000 Has Ex Left-Turn Lane 0.223 0.000 Pedestrian Presence -0.218 0.000 Number of Signals per Mile -0.217 0.000 Control Delay per Mile -0.210 0.000 Speed Limit 0.198 0.000 Median Presence 0.175 0.000 Stops per Signal -0.159 0.000 Lane Marking Quality 0.110 0.000 Median Width 0.107 0.000 Variation in Speed -0.084 0.000 Number of Through Lanes -0.065 0.003 Separation Between Sidewalk and Travelway 0.055 0.010 Exhibit 65. Correlation Between Explanatory Variables and LOS Ratings. width. Two-way left-turn lanes do not count as exclusive left-turn lanes. Trees = 3 if many, 2 if some, 1 if few or none Pavement Quality = 4 if new, 3 if typical, 2 if cracked, 1 if poor. The R-square statistic for this model equals 0.673, mean- ing 67 percent of the variation in mean participant ratings can be estimated by the model; however, several variables included in the model do not contribute significantly to the overall model predictive power, as indicated by their high p-values. To address the inclusion of variables that are not statistical contributors to the model, another regression model was developed that included only the number of stops and the presence of an exclusive left-turn lane, those variables which were significant in Model 1. This new model’s details are pro- vided in Exhibit 68. In this model each of the two variables, number of stops, and presence of an exclusive left-turn lane were significant at the 0.05 level resulting in the following: Mean Auto LOS = 4.327 − 0.622 (Stops) + 0.293 (Left Turn Lane) (Eq. 10) Where Mean Auto LOS = 6.0 for LOS A and 1.0 for LOS F Stops = Number of times in video clip that auto speed drops below 5 mph. Variable tau Rank Correlation Coefficient Space Mean Speed 1.00 Total Travel Time 0.617 Lane Width -0.694 Number of Stops per Mile 0.270 Sign Quality 0.442 Tree Presence 0.474 Sidewalk Width -0.423 Has Ex Left-Turn Lane 0.264 Pedestrian Presence -0.445 Number of Signals per Mile -0.270 Control Delay per Mile -0.721 Speed Limit 0.381 Median Presence 0.147 Stops per Signal -0.462 Lane Marking Quality 0.111 Median Width 0.117 Variation in Speed -0.287 Number of Through Lanes -0.222 Separation Between Sidewalk and Travelway 0.104 Exhibit 66. Correlation Between Space Mean Speed and Other Explanatory Variables.

64 Variable Standard Coefficient t-Statistic Statistical Significance (p-value) Constant 3.8 9.832 0.00* Number of Stops -0.530 -4.154 0.00* Median Presence -0.155 -0.898 0.377 Presence of Ex. Left-Turn Lane 0.355 1.903 0.067* Tree Rating 0.098 0.816 0.421 Pavement Quality 0.205 1.556 0.130 *Statistically Significant at the 0.10 confidence interval Exhibit 67. Multiple Linear Regression Model #1. Left-Turn Lane = Presence of exclusive left-turn lane at all intersections. Equals one if present, zero otherwise. Exclusive left-turn lane can be of any length or width. This second model has an adjusted R-square value of 0.647, slightly inferior to Model 1. Similar model formats were attempted using average space mean speed as the primary predictor of mean participant rat- ing; however, the adjusted R-Square values were lower at 0.545, meaning only 54.5% of the variation in mean partici- pant ratings could be explained by the model. For this model, explanatory variables space mean speed, presence of an ex- clusive left-turn lane, and tree rating were included in the stepwise regression analysis. In this case, tree rating did not contribute significantly to the prediction power of the model and so was removed. Exhibit 69 contains statistical informa- tion for Model 3. The end result is a model of the following form Mean Auto LOS = 2.673 + 0.479 (Speed) + 0.403 (Left-Turn Lane) (Eq. 11) Where Mean Auto LOS = 6.0 for LOS A and 1.0 for LOS F Speed = average space mean speed in mph. Left-Turn Lane = Presence of exclusive left-turn lane at all intersections. Equals one if present, zero otherwise. Exclusive left-turn lane can be of any length or width. This model has an adjusted R-square value of 0.545, infe- rior to Models 1 and 2. The lowest value that this model can produce is 2.673, the value of the constant. Thus the LOS could never be below LOS E. Limitations of Linear Regression Modeling Linear regression techniques are not particularly the best model specification choice when modeling ordered response variables in that linear regression models attempt to deter- mine the best-fitting linear equation according to the least- square criterion, such that the sum of the squared deviations of the predicted scores from the observed scores is minimized to give the most accurate prediction. This assumes that, for a measured change in the explanatory variables, there is a measured linear change in the dependent variable, namely the mean participant rating. Linear regression models also predict a continuous variable, which is different than what was asked of participants in the study. For example, linear re- gression models will also predict values such as 3.42 LOS, lying between LOS C and D. These limitations led the research team to investigate the use of cumulative logistic regression, which can predict the Variable Standard Coefficient t-Statistic Statistical Significance (p-value) Constant 4.327 16.428 0.00* Number of Stops -0.622 -5.152 0.00* Presence of Ex. Left-Turn Lane 0.293 2.427 0.021 *Statistically Significant at the 0.05 confidence interval Exhibit 68. Multiple Linear Regression Model #2. Variable Standard Coefficient t-Statistic Statistical Significance (p-value) Constant 2.673 10.483 0.00* Space Mean Speed 0.479 3.657 0.01* Presence of Ex. Left-Turn Lane 0.403 3.075 0.004* *Statistically Significant at the 0.05 confidence interval Exhibit 69. Multiple Linear Regression Model #3.

65 probability of responses within each LOS based on a combi- nation of explanatory variables. This characteristic also allowed the research team to make use of the nearly 1,650 observations contained in the modeling database rather than just 35 mean estimates of LOS (1 mean LOS for each video clip shown), and to predict the discrete outcome (i.e. 1,2, . . ., 6) as generated from the video lab surveys. Cumulative Logistic Regression For the auto LOS survey, the overall ratings (RatingNum) have a hierarchical ordering that varies from 1 (worst rating, or LOS F ) to 6 (best rating, or LOS A). The discrete nature of RatingNum rules out the use of ordinary Linear Regression, because it requires the response to be a continuous variable. Cumulative logistic regression addresses the issue of model- ing discrete variables with hierarchical ordering. Consider the following cumulative probability P(Y ≤ j | x) and define the logistic model for this probability as (Eq. 12) In general, P(Y = j | x) = 1 − P(Y ≤ j − 1 | x), so estimated probabilities for all scores can be obtained. The vector β′ rep- resents the vector of coefficients for both LOS ranges (there are 6 − 1 = 5 such intercept coefficients designated as α′s) as well as the coefficients of the independent variables consid- ered in the model (designated as β′s). Equation 12 can be rewritten as (Eq. 13) Each cumulative probability has its own intercept αj; the values of αj are increasing in j since P(Y ≤ j | x) increases in j for fixed x. The model assumes the same effects βtree_presnce, βstops_per_milet and βPres__Of_Ex__LT_Lane for each j. In order to have an appropriate interpretation of the inter- cept values, consider Model 3 for two scores j and k with j < k and assume values = 0 for the two dummies tree_presence and Pres__Of_Ex__LT_lane. After some algebraic manipula- tions we have P(Y ≤ k | stops_per_mile)= P(Y ≤ j | stops_per_mile + (αk − αj) / β). (Eq. 14) Cumulative probability for j is the same as the cumulative probability for k but evaluated at a stops_per_mile value dis- placed by an amount dependent on the positive difference be- tween intercepts at score j and k, and the parameter β. Exhibit 70 illustrates this model with increasing values of each αj and positive value of β for stops_per_mile. The model coefficients, estimated using the Maximum Likelihood estima- tion methods, and their significance are shown in Exhibit 71. P Y j( | ) exp( ) exp( ) ≤ = + x x x (3) β β ′ ′1 Ln P Y j x P Y j x x ( | ) ( | ) ( ) ≤ − ≤ = 1 β′ What the increasing value of the intercept guarantees in this case is that for each (integer) value of RatingNum the se- quence of cumulative probabilities for a certain value l of stops_per_mile are in the right order, meaning that, (Eq. 15) A positive slope is evident in Exhibit 70 as the increment in the cumulative probability for a particular RatingNum score when stops_per_mile value increases. The difference between successive curves for RatingNum scores determines the prob- ability P(Y = j | stops_per_mile) for an individual RatingNum score given a fixed value of stops_per_mile. For instance, the value of P(Y ≤ 1) = P(Y = 1) is higher when stops_per_mile = 18 than when stops_per_mile = 1 and the value of P(Y = 5) = P(Y ≤ 5) − P(Y ≤ 4) is higher when stops_per_mile = 1 than for stops_per_mile = 18, so it appears that, with higher prob- ability, high ratings in LOS are given to trips with fewer stops per mile. Also shown in the exhibit are the marginal proba- bilities of the various levels of service (A to F) when the num- ber of stops per mile is fixed at 2. Best Candidate Auto LOS Models Preliminary modeling analysis has resulted in two fairly strong models, one which uses number of stops per mile and the other which uses average space mean speed as the primary explanatory variable. Both models perform well. Both mod- els are presented with “trees” or “no trees” options. The rec- ommended models are shown in Exhibits 72 through 74. The Pearson Correlation measures the ability of each model to reproduce the observed video clip ratings of level of service. A higher value indicates a better model fit. The high- est possible value is 100%. The Akaike Information Criterion (AIC) represents how close fitted values are to actual values taking into account the number of parameters included in each model (Agresti, 2002 [97]). Lower values indicate a superior model of the data. Model 4 (stops per mile, presence of an exclusive left-turn lane and presence of trees) reported the lowest AIC measure and the highest Pearson correlation. Performance of Candidates A preliminary analysis of the ability of Models 4, 5, and 6 to predict the distribution of ratings of LOS as reported by participants was also undertaken. (The performance of Model 7 is presented in a later section describing refinement options for the recommended Auto LOS Model.) P Y stops by mile l P Y stops by mile( | _ _ ) ( | _ _≤ = ≤ ≤ =1 2 l P Y stops by mile l P Y stops by m ) ( | _ _ ) ( | _ _≤ ≤ = ≤ ≤3 4 ile l P Y stops by mile l = ≤ ≤ = ≤ ) ( | _ _ )5 1

66 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 0 2 4 6 81 3 5 7 9 10 11 12 13 14 15 16 17 18 Stops by mile (x) P(Y≤j) P(Y≤1) P(Y≤2) P(Y≤3) P(Y≤4) P(Y≤5) P(Y=1| x=2) P(Y=2| x=2) P(Y=3| x=2) P(Y=4| x=2) P(Y=5| x=2) P(Y=6| x=2) Parameter DF Estimate Standard Erro r Wald Chi-Square Pr > ChiSq Intercept 1 1 -2.9189 0.2270 165.4053 <.0001 Intercept 2 1 -1.8273 0.2075 77.5198 <.0001 Intercept 3 1 -0.8529 0.2009 18.0246 <.0001 Intercept 4 1 0.2832 0.2005 1.9951 0.1578 Intercept 5 1 2.0937 0.2091 100.3006 <.0001 stops_per_mile 1 0.2033 0.0184 122.3357 <.0001 Pres__Of_Ex__LT_Lane 1 -0.5218 0.1111 22.0627 <.0001 Tree_Presence 1 -0.3379 0.0612 30.4761 <.0001 Parameters for Cumulative Regression Model Applied to Auto LOS—Stops per Mile Model– Model 4 Exhibit 71. Maximum Likelihood Estimates for Model #4. Exhibit 70. Example Cumulative Logit Distribution of LOS. Because the strength of the cumulative probability models lies in their ability to predict the distribution of LOS ratings for a particular combination of explanatory variables, the models were tested to determine their ability to accurately predict a distribution of responses as compared with those collected at the four study sites. One-third of the auto re- sponse dataset was reserved to test the fit of various models developed in this study and was not used in the model cali- bration. A subset of four clips was chosen to be shown in each of the four study sites in Phase II of the study. These four clips resulted in the largest number of observations in the test dataset, so only these four clips have been tested with Models 4, 5, and 6. The remaining clips in the test database do not have enough observations to develop a robust distribution of response data. Clips 2, 15, 52, and 56 are discussed in the fol- lowing analysis. Exhibits 75 through 78 compare the observed LOS rating distributions to those predicted by Model 4 (Stops/Mile; left- turn lane presence; tree presence index), Model 5 (Space Mean Speed; median presence index; tree presence index)

67 Parameter Model 4 Model 5 Model 6 Model 7 Intercept LOS E, Alpha1= -2.92 -0.73 -3.80 -1.19 Intercept LOS D, Alpha2= -1.83 0.28 -2.70 -0.20 Intercept LOS C, Alpha3 = -0.85 1.21 -1.74 0.71 Intercept LOS B, Alpha4= 0.28 2.32 -0.62 1.80 Intercept LOS A, Alpha5= 2.09 4.16 1.16 3.62 Stops Per Mile, Beta1 = 0.20 0.25 Presence of Left-Turn Lanes, Beta2 = -0.52 -0.34 Mean Speed (mph), Beta1 = -0.063 -0.084 Median Presence (0-3), Beta2 = -0.33 -0.22 Presence of Trees, Beta3 = -0.34 -0.42 Pearson Correlation= 79% 76% 77% N/A Akaike Information Criterion (AIC) 4944.0 5034.1 5022.8 5076.4 Exhibit 72. Recommended Auto LOS Models. Clip Number Number of Stops/Mile Average Space Mean Speed (mph) Median Presence (0-No 1-One-way Pair 2-TWCLTL 3-Raised Ex. Lt.-Turn- Lane Presence (0-No 1-Yes) Tree Presence (1-Few 2-Some 3-Many) HCM LOS 2 0 34.5 3 1 2 A=6 15 6 7.86 3 1 1 F=1 52 7 7.9 0 0 1 E=2 56 2 23.1 3 1 3 C=4 Exhibit 73. Test Clip Characteristics. Models Compared Pearson Correlation Coefficient HCM LOS to Mean Observed LOS 0.465 Mean Observed LOS to Mean LOS – Model 4 0.787 Mean Observed LOS to Mean LOS – Model 5 0.764 Mean Observed LOS to Mean LOS – Model 6 0.770 Exhibit 74. Correlation Coefficients of Auto LOS Models. Clip 2 - Comparing LOS Distributions of Test Data and Models 4, 5 and 6 (N=59) - HCM LOS=6 0.00% 5.00% 10.00% 15.00% 20.00% 25.00% 30.00% 35.00% 40.00% 45.00% 50.00% 1 2 3 4 5 6 LOS Ratings (A=6; F=1) Pr ob ab ili ty M as s Fu nc tio n TEST MODEL 4 MODEL 5 MODEL 6 Exhibit 75. Evaluation of Models Against Clip 2 Ratings.

68 TEST MODEL 4 MODEL 5 MODEL 6 Clip 15 - Comparing LOS Distributions of Test Data and Models 4, 5 and 6 (N=42) - HCM LOS=1 0.00% 5.00% 10.00% 15.00% 20.00% 25.00% 30.00% 35.00% 1 2 3 4 5 6 LOS Rating (A=6 F=1) Pr ob ab ili ty M as s Fu nc tio n Exhibit 76. Evaluation of Models Against Clip 15 Ratings. and Model 6 (Stops per mile; left-turn lane presence) for the four video clips. Exhibit 73 lists the conditions depicted in each of the four clips and the current HCM-estimated LOS. Overall the models appear to track comparatively well with each other and with the data, in that there is a general increase/ decrease in the estimation of LOS probability. Model 4 has slightly higher predictive power—it tends to track slightly closer to the test dataset represented by the periwinkle bar. For Clip 15, there is a definite difference between the HCM LOS of F and the distribution of LOS as provided by the study participants, which is shifted toward the right, meaning higher LOS ratings. In this clip, there are many stops along a short ar- terial, however, there is only low to moderate traffic congestion so that the test vehicle is always in the first position of the queue TEST MODEL 4 MODEL 5 MODEL 6 Clip 52 - Comparing LOS Distributions of Test Data and Models 4, 5 and 6 (N=44) - HCM LOS=2 0.00% 5.00% 10.00% 15.00% 20.00% 25.00% 30.00% 35.00% 1 2 3 4 5 6 LOS Rating (A=6 F=1) Pr ob ab ili ty M as s Fu nc tio n Exhibit 77. Evaluation of Models Against Clip 52 Ratings.

69 TEST MODEL 4 MODEL 5 MODEL 6 Clip 56 - Comparing LOS Distributions of Test Data and Models 4, 5 and 6 (N=50) - HCM LOS=4 0.00% 20.00% 40.00% 60.00% 10.00% 30.00% 50.00% 1 2 3 4 5 6 LOS Rating (A=1; F=6) Pr ob ab ili ty M as s Fu nc tio n Exhibit 78. Evaluation of Models Against Clip 56 Ratings. at each signal. In addition, it is a relatively clean and newly built out area in the Washington, DC, suburb of Arlington, VA. This combination of factors may have led participants to rate the video higher, despite the overall low space mean speed of 7.86mph and the high number of stops (6/mile). Comparing the mean LOS as observed, the HCM LOS, and the model performance for Models 4, 5, and 6 we find that the HCM overall tended to underpredict the mean LOS as observed in the video laboratories. Model 6 also tends to underpredict mean observed LOS. A correlation analysis of the various models and the test dataset shows that Models 4, 5, and 6 all have superior corre- lation to the mean video clip ratings than the HCM (see Exhibit 74). Essentially the models all show a positive correlation, in that the compared models track the same in the positive or negative direction. The current HCM LOS method can only explain approximately 46 percent of the variation in mean observed LOS ratings. The three models developed for this study all perform much better and can explain, on average, approximately 75 percent of the variation in mean observed LOS ratings. The best fitting model is Model 4, which uses stops per mile, presence of an exclusive left-turn lane and the presence of trees to estimate the observed LOS ratings. 5.2 Recommended Auto LOS Model The recommended auto LOS model (Model 6 above) pre- dicts the average degree of satisfaction rating for the facility, where LOS A is “very satisfied” and LOS F is “very dissatisfied.” AutoLOS = Mean (LOS) (Eq. 16) The average or “mean” LOS rating is the sum of the prob- abilities of an individual giving a facility a given LOS rating multiplied by the numerical equivalent of that LOS rating (J) (worst = 1, best = 6). (Eq. 17) Where J = 1 for the worst LOS rating and 6 for the best LOS rating. The Mean LOS number is converted to a mean letter grade for the facility according to Exhibit 79. The numerical thresholds for converting the mean score to the mean LOS letter grade differ from the scores (J) used to compute the mean score. Section 4.7 explained why different thresholds are used to convert the mean result to a letter grade. The probability that a person will rate a given facility as exactly LOS J is computed by subtracting the cumulative probability of Mean LOS LOS J J J ( ) Pr( )*= = = ∑ 1 6 LOS Mean Numerical Score A ≤ 2.00 B >2.00 and ≤ 2.75 C >2.75 and ≤ 3.50 D >3.50 and ≤ 4.25 E >4.25 and ≤ 5.00 F > 5.00 Exhibit 79. Auto LOS Thresholds for Mean Numerical Scores.

70 giving the facility a lower LOS rating from the cumulative prob- ability of giving the facility a LOS J rating or worse. Pr(LOS = J) = Pr(LOS ≤ J) − Pr(LOS ≤ J − 1) (Eq. 18) The probability that a person will rate a given facility as LOS J or worse is given by the ordered cumulative logit model as shown below: (Eq. 19) Where Pr(LOS<=J) = Probability that an individual will respond with a LOS grade J or worse. exp = Exponential function. αJ = Alpha, Maximum numerical threshold for LOS grade J (see Exhibit 80). βK = Beta, Calibration parameters for attributes (see Exhibit 80). XK = Attributes (k) of the segment or facility (see Exhibit 80). Two ordered cumulative logit models are recommended, both using the same form. Model 1, derived from a statistical analysis of the video lab data, predicts auto LOS as a function of the number of stops per mile and the presence of exclusive left-turn lanes. Model 2 was created to provide a speed-based model op- tion for auto LOS. Model 2 predicts auto LOS as a function of the percent of free-flow speed and the type of median. The parameters for this model were first derived statistically from the video lab data. The resulting model, however, did not produce a full LOS A to LOS F range of results for the streets in the video lab sample. Given that public agencies may be re- luctant to adopt a LOS model that cannot predict LOS A, the LOS intercept values for the model were modified manually to obtain a full LOS range of results for the streets in the video clip sample while attempting to maintain as high a match per- centage with the video lab results as possible. Pr( ) exp( )( ) LOS J xJ k k k ≤ = + − −∑ 1 1 α β Model 1 provides a greatly superior statistical fit with the video lab data, however; this model does not produce LOS A for the streets contained in the video lab sample. Model 2 provides an inferior statistical fit with the data, but provides numerous LOS A results for the streets in the video lab sam- ple. Both models predict LOS F for one or more of the streets in the video lab sample. The attribute, stops per mile, is the number of times a ve- hicle decelerates from a speed above 5 mph to a speed below 5 mph, divided by the length of the urban street segment under consideration. The attribute, Left-Turn-Lane Presence, takes on the fol- lowing values: • 1 if exclusive left-turn lane at intersections, • 0 if not. If the exclusive left-turn lanes do not provide sufficient storage for left-turning vehicles, then the number of stops per mile would be affected, which would, in turn, adversely affect the perceived level of service. The attribute, Percent Speed Limit, is the ratio of the actual average speed (distance traveled divided by the average travel time for the length of the arterial including all delays) to the posted speed limit for the street. The attribute, Median Type, is equal to • 0 if no median, • 1 if one-way street, • 2 if a painted median is present, and • 3 if a raised median is present. The threshold values, αj, and the attribute equation coeffi- cients, βk, of the ordered cumulative logit function are calibrated using the maximum likelihood estimation (MLE) process applied to paired data of facility characteristics and perceived LOS collected from people participating in the video laboratory surveys. Parameter Model 1 Model 2 Alpha Values Intercept LOS E = -3.8044 1.00 Intercept LOS D = -2.7047 2.00 Intercept LOS C = -1.7389 2.50 Intercept LOS B = -0.6234 3.00 Intercept LOS A = 1.1614 4.00 Beta Values Stops/Mile= 0.2530 N/A Left-Turn-Lane Presence (0-1), = -0.3434 N/A Percent Speed Limit N/A -5.74 Median Type (0,1, 2, 3) N/A -0.39 Exhibit 80. Alpha and Beta Parameters for Recommended Auto LOS Models.

71 5.3 Performance of Auto LOS Models Exhibit 81 shows how the mean LOS values produced by the recommended auto LOS model compare with the mean LOS values reported by the video lab participants. The HCM-predicted LOS is included as well. The HCM matched the video labs 26% of the time, while the proposed auto LOS Model 1 (stops) matched the video labs 69% of the time. The alternate proposed auto LOS Model 2 (speed) matched 37% of the video clip mean LOS ratings. Model 2 ranges from LOS A to F, while Model 1 ranges from LOS B to F. Model 1 fits the video clip data better but does not achieve LOS A for the streets in the video clip sam- ple. Model 2 has a significantly poorer fit with the data (but still better than the HCM); however, it does produce LOS A for nine of the streets in the video clip sample. Several different sections or time periods of the same street were used for many of the clips. 61 Rt 50 1 50 28 1.4 100% 0.00 A C B C 56 Sunset Hills Rd 2 40 23 2.0 100% 3.00 A C B A 2 Gallows Road 3 35 35 0.0 100% 3.00 B A B A 65 Lee Hwy 2 40 36 0.0 100% 2.00 B A B A 63 Rt 50 1 50 42 0.0 100% 3.00 B A B A 5 Wilson Blvd 3 35 30 0.0 100% 3.00 B B B A 62 Rt 50 1 50 37 0.0 100% 0.00 B B B A 13 Washington Blvd 3 35 25 0.0 0% 0.00 B B B A 7 Wilson Blvd 3 35 20 0.0 100% 1.00 B C B B 54 Lee Hwy 2 40 25 3.3 100% 2.00 B C B A 53 Prosperity 2 40 19 1.7 100% 3.00 B D B B 6 Clarendon 3 35 18 2.3 100% 1.00 B D B B 10 Washington Blvd 3 35 17 3.8 0% 0.00 B D C C 20 Rt 50 1 50 16 1.8 100% 3.00 B E B C 64 Rt 50 1 50 20 2.0 100% 3.00 B E B B 58 Sunrise Valley Rd 2 40 11 1.7 100% 3.00 B F B C 1 Rt 234 1 50 15 2.0 100% 3.00 B F B C 29 Rt 234 2 40 23 2.0 100% 3.00 C C B A 19 23rd St 4 30 16 5.8 0% 0.00 C C C C 12 Wilson Blvd 3 35 14 4.3 0% 0.00 C D C D 60 Lee Hwy 2 40 15 2.0 100% 2.00 C E B C 21 Rt 50 1 50 20 4.0 100% 3.00 C E C B 8 Wilson Blvd 3 35 14 4.1 100% 1.00 C E C C 52 M St 4 30 8 7.3 0% 0.00 C E D E 55 Braddock Rd 2 40 13 2.2 100% 3.00 C F B C 59 Sunset Hills Rd 2 40 12 4.9 0% 0.00 C F C E 15 Glebe Road 2 40 8 6.0 100% 3.00 C F C D 14 Glebe Road 2 40 11 6.0 100% 3.00 C F C C 57 Sunset Hills Rd 2 40 17 3.3 0% 0.00 D D C D 16 Fairfax Drive 3 35 12 7.3 100% 3.00 D F C C 51 M St 4 30 7 9.1 0% 0.00 D F D E 25 M St 4 30 11 3.7 0% 0.00 E D C D 23 M St 4 30 8 5.6 0% 0.00 E E C E 30 M St 4 30 7 14.5 0% 0.00 F F F E 31 M St 4 30 4 18.0 0% 0.00 F F F F % Exact Match To Video 100% 26% 69% 37% % Within 1 LOS of Video 100% 46% 94% 89% Art Spd Lim Actual Stops Left Ln Med Video HCM Model Model Clip # Street Class (mph) (mph) (stps/mi) (%) (1,2,3) LOS LOS 1 LOS 2 LOS Exhibit 81. Evaluation of Proposed Auto LOS Models.

Next: Chapter 6 - Transit LOS Model »
Multimodal Level of Service Analysis for Urban Streets Get This Book
×
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

TRB’s National Cooperative Highway Research Program (NCHRP) Report 616: Multimodal Level of Service Analysis for Urban Streets explores a method for assessing how well an urban street serves the needs of all of its users. The method for evaluating the multimodal level of service (MMLOS) estimates the auto, bus, bicycle, and pedestrian level of service on an urban street using a combination of readily available data and data normally gathered by an agency to assess auto and transit level of service. The MMLOS user’s guide was published as NCHRP Web-Only Document 128.

Errata

In the printed version of the report, equations 36 (pedestrian segment LOS) and 37 (pedestrian LOS for signalized intersections) on page 88 have been revised and are available online. The equations in the electronic (dpf) version of the report are correct.

In June 2010, TRB released NCHRP Web-Only Document 158: Field Test Results of the Multimodal Level of Service Analysis for Urban Streets (MMLOS) that explores the result of a field test of the MMLOS in 10 metropolitan areas in the United States.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!