Chapter 9. Adding variables 1. VIEDA for the intersection data. In problem 2 of Chapter 8 the model equation 𝛽0 × (𝑀𝑎𝑗𝑜𝑟 𝑅𝑜𝑎𝑑 𝐴𝐴𝐷𝑇)𝛽1 × (𝑀𝑖𝑛𝑜𝑟 𝑅𝑜𝑎𝑑 𝐴𝐴𝐷𝑇)𝛽2 was fitted to the multivehicle injury crash counts of the Edmonton intersection data in the ‘Edmonton 4-legged Signalized Intersections.xls’ file. A part of that file is shown in Figure 1. Figure 1 The Negative Binomial fit of the model equation to the data is in the ‘Edmonton NB.xls’ file. Not used in the model equation are the variables ‘Area’ (in column B), ‘Number of approaches with left-turn lane, lt’ (in column C), and ‘Number of approaches with right-turn lane, rt’ (in column D). The question is whether the ‘Area’ and ‘lt’ variables are safety related and, if yes, how might they be represented in the model equation. 2. What was gained by adding the ‘Area’ variable By maximizing the NB likelihood with 𝛽0 × (𝑀𝑎𝑗𝑜𝑟 𝐴𝐴𝐷𝑇)𝛽1 × (𝑀𝑖𝑛𝑜𝑟 𝐴𝐴𝐷𝑇)𝛽2 as the model equation the parameter estimates in Figure 2 were obtained. Figure 2 With separate models for ‘Urban’ and ‘Suburban’ intersections, the parameter estimates were those in Figure 3. Figure 3 Was anything gained? a. Is the change in log-likelihood justified in view of the AIC and BIC? b. Compare CURE plots The C-F spreadsheets for ‘All’, ‘Urban’ and ‘Suburban’ intersections are in the ‘Edmonton C-Fs.xls’ file.