1. Transportation Planning Asian Institute of Technology
Trip Generation 2
Transportation Planning
Asian Institute of Technology

2. Contents T-Statistics Linear Regression Analysis Concerns
Trip Rate Method
Category (Cross Classification) Analysis

3. T-Statistic
Used to test whether variables have high significance to the model
Hypothesis
Two-tailed test
If T-statistic is between ± tcritical
accept H0: a and/or b is not significant to the model.
If T-statistic is not between ± tcritical
reject H0: a and/or b is significant to the model.

4. T-Statistic
T-test for parameter a
T-test for parameter b

5. T-Statistic
Linear Regression

6. T-Statistic
T-test for a at 95% level of confidence

7. T-Statistic
Linear Regression

8. T-Statistic
T-test for b at 95% level of confidence

9. Linear Regression Analysis: Concerns
Relationship between independent and dependent variables may not be linear.
Parameters may change over time.
Independent variables may have correlations.
Model may not be logical.
Zonal average may not reflect characteristics of the analysis zones.

10. Linear Regression Analysis: Concerns
Zonal-Based Multiple Regression
A method to find a linear relationship between the number of trips produced or attracted by zone and average socioeconomic characteristics of the households in each zone.

11. Linear Regression Analysis: Concerns
Zonal-Based Multiple Regression
Interesting considerations include:
Zonal models can only explain variation in trip making behavior between zones. Zones should have homogenous socioeconomic composition and represents a wide range of conditions.
Intercepts are often expected to be zero, i.e. the regression line passes through the origin.
“Null” zones do not contain information about certain independent variables.
Zonal total vs zonal means. The analyst must decide between using aggregate quantity (cars per zone, trips per zone) or zonal means (rate) (cars per household per zone)

12. Linear Regression Analysis: Concerns
Zonal-Based Multiple Regression
Zonal Totals
The regression equation
Zonal means
where yi = Yi/Hi; xi = Xi/Hi; ei = Ei/Hi
Hi = the number of households in zone i

13. Linear Regression Analysis: Concerns
Zonal-Based Multiple Regression
Interesting notes
The magnitude of error depends on the size of the zone. The larger zone size means larger error.
Aggregate variables help improve R2.

14. Linear Regression Analysis: Concerns
Household-Based Multiple Regression
Smaller zones reduce intra-zonal variation, but require more budget and involve higher risk in sampling errors.
Household-based survey is independent of zone boundary.
Each household’s characteristics and travel behavior are observed.

15. Linear Regression Analysis: Concerns
Example 3-1 Consider trips per household (Y), number of workers (X1), and number of cars (X2), assuming large enough sample size so the regressor can be estimated as t- student distribution.
Step
Equation R2 1
Y = 2.36X1
0.203 2
Y = 1.80X X2
0.325 3
Y = X X2
0.384 t
(3.7) (8.2) (4.2)

16. Linear Regression Analysis: Concerns
Example 3-1
t values should be compared whether t > ta;d (one tailed test) as both dependent variables have positive effects to the number of trips.
The intercept 0.91 is not large as compared to 1.44 times the number of worker and 1.07 times the number of cars.
H0 is rejected in all cases.
The goodness of fit of the model can be validated by comparing the observed and modeled trips by group.

17. Linear Regression Analysis: Concerns
The Problem of Non-linearities
Two methods to corporate variables with non-linear relationship into the linear model.
Transform the variables to linearize the effect (e.g. take log, take ln, square root or raise to the power). This could take time and effort as there are no established rule.
Use dummy variables by discrete intervals and treat each of the dummy variables separately in the model.

18. Linear Regression Analysis: Concerns
The Problem of Non-linearities
Applying dummy variables:
Z1 = 1 for the household with 1 car and otherwise
Z2 = 1 for the household with 2 or more cars and 0 otherwise
Model 3 in the previous example becomes:
Y = X Z Z2
2 or more cars
1 car
0 car
HB Trips per household (Y)
Number of workers (X1)

19. Trip Rate Method
Analysis of trip generation (production or attraction) per unit of land use factor.
Factors generating trips
Residential – residence, dwelling units, total area, etc.
Commercial – employments, parking spaces, gross floor area, gross leasable area, etc.
Industrial – factory floor area, workers, etc.
Equivalent to linear equation with one variable and go through origin (0,0)

20. Trip Rate Method Trip Rate Manuals
Institute of Transport Engineers (ITE) Trip Generation Manual shows trip rates of various land use.
Dubai Trip Generation and Parking Rates Manual uses similar form of ITE manual but adds parking data.
Bangkok/Thailand performs some trip generation in some feasibility study, but do not have an official manual.

21. Trip Rate Method Components in Trip Rate Manual Type of Traffic
Average Rate/Range of Rate
Standard Deviation
Number of Studies
Average Number of Trips and Quantity of Unit (upper/lower limits)
Equation of Fitted Curve
Coefficient of Determination (R2)

22. Trip Rate Method
ITE Trip Generation Manual (1)

23. Trip Rate Method
ITE Trip Generation Manual (2)

24. Trip Rate Method
ITE Trip Generation Manual (3)

25. Trip Rate Method Other Trip Generation Manuals
San Diego Municipal Code for Trip Generation Manual
Dubai Trip Generation and Parking Rates Manual

26. Category Analysis
Group similar socioeconomic characteristics together.
Trip Production
Trip generated from households or other trip generation units
Trip rate directly varies by status/characteristics of the household
Independent variables used for classification – vehicle ownership, household, average income, etc.
Trip Attraction
Classified by employment characteristics such as store, office and factory
Independent variables used for classification – density

27. Category Analysis Example 4-1 Trip Rate Households Vehicle ownership
Household size 1 2 3 4 5
0.57
2.07
3.30
4.57
6.95
1.45
3.02
3.51
5.52
7.90
1.82
3.39
3.75
5.89
8.27
Vehicle ownership
Household size 1 2 3 4 5
100
200
150 20
300
500
210 50 80 60

28. Category Analysis Example 4-1
Households with 1 person and 0 car will make = 0.57 x 100 = 57 trips/day
Households with 1 person and 1 car will make = 1.45 x 300 = 435 trips/day
Households with 2 persons and 0 car will make = 2.07 x 200 = 414 trips/day
Etc.

29. Cross Classification or Category Analysis
Advantages
Independent of zone system of the study area.
No prior assumptions about the shape of the relationship
Relationship can differ inform from class to class

30. Cross Classification or Category Analysis
Disadvantages
The model does not permit extrapolation beyond its calibration strata.
No goodness of fit measures for the model.
Large samples are required to fulfill the minimum sample size in each category.
There is no effective way to choose the best variables. The model is not rational explainable.
An addition to strata will increase the sample group and sampling effort enormously.

31. Cross Classification or Category Analysis
Improvements to the basic model Multiple Classification Analysis (MCA) uses mean row and column values to estimate the average trip rates. (Compare with Table 4.9) Overall mean trip rate = 1.54
HH Size
0 car
1 car
2+ car
Deviations
1 person
0.00
0.46
1.37
-1.07
2 or 3 persons
1.27
2.18
-0.26
4 persons
1.05
1.85
2.76
0.32
5 persons
1.09
1.89
2.80
0.36
-0.81
-0.01
0.90

32. Assignment #4
From last assignment, prove that a and b are not zero