Transport Demand - Modal Split

Question 1

What is modal spilt?

Answer 1

By which mode will trips be undertaken?

Question 2

What does modal spilt vary by?

Answer 2

zone
purpose
time

Question 3

What factors influence mode choice?

Answer 3

Characteristics of Trip maker
Characteristics of the journey
Characteristics of the transport system
Characteristics of trip origin and destination

Question 4

What is the factor based approach?

Answer 4

Uniform factoring:
Tij ^k = F^k Tij
where F^k = proportion of trips by mode k

Sector Based factoring:
Tij ^k = F^k Tij
where F^k = proportion of trips by mode k between zones i

Question 5

What are the pros and cons of factor based approach?

Answer 5

Pros:
Simple to develop and apply
suitable for areas with no or little likelihood of significant change to existing mode choice behaviours

Cons:
ignores all factors influencing mode choice
how should these factors change for the forecast years

Question 6

What is modal penalty?

Answer 6

The penalty associated with the preference of one mode over another (Even though there is a difference in cost)

Question 7

What is the binary logit model?

Answer 7

Pij^c = exp (-bCij^c)/(exp(-bCij^c) +exp (-b(Cij^B +d)))

Question 8

How is calibrating a binary logit model?

Answer 8

By using Ln(P^C / 1- P^c)) = b (C^B - C^C) + bd

in the form y = mx +c

Question 9

What is a Multinomical logit model?

Answer 9

When more than two modes are used -
P^1 = proportion of trips by mode 1 = exp (-bC^1) /∑(exp (-bC^N))

N way structure
Assumes alternatives have equal weight

Question 10

What is hierarchical logit model?

Answer 10

Modes with common elements are combined into composite modes

Relies on composite costs which depart further from reality as one moves up the hierarchy

Question 11

What is the pivot point model?

Answer 11

P^1 ij = proportion of trips by mode 1 from zone i to zone j
= Tij / (1+exp(aC^2 ij - bC^2 ij + mij)

a and b are modal constants

m = combined socio economic variable
m = ∑ wx = ∑(weight x socio economic variable)

must have pre existing matrix for each mode
must have skim matrice for each mode
find a and be and ma through iterative process

useful for estimating the impact of incremental changes in modal costs

Question 12

What are direct demand models?

Answer 12

Combines trip generation, distribution and mode split into a single well defined model:
T^k ij = f (origin attributes, destination attributes, LOS)

Question 13

What are the pros of direct demand models?

What are the cons?

Answer 13

Overcomes theoretical problems of 4 step process
proved useful inter-urban studies

Cons:
large number of parameters to calibrate
requires lots of input data
few successful urban applications

Question 14

What is a discrete choice model?

Answer 14

Aggregate modesl based on observations of group behaviour
Probability of an individual choosing a given option = f

Attractiveness = utility
utility usually defined as a linear combination of key variables

Question 15

What are the four different model used for modal spilt?

Answer 15

Discrete Choice model
Direct Demand model
Pivot Model
Logit Model
S-shaped mode split curves
factor based models

Question 16

What is utility?

Answer 16

the perceived ability of something to needs or wants

represents satisfaction experience by the consumer of a good

Question 17

What is the utility maximisation principle?

Answer 17

People choose the alternative that has the highest perceived utility for them
Since the perceived utilities are not exactly known, the choice models incorporate an error term that accounts for the stochasticity of choices

Question 18

What does the assumptions of logit model?

Answer 18

Assume that error terms are independent, identically and Gumbel distributed

Question 19

What is maximum likelihood estimation (MLE)

Answer 19

• We use MLE to estimate the coefficients of a utility function
• MLE searches among all possible values of utility coefficients and finds the ones that best describe the choice observations

Question 20

How is maximum likelihood calculated?

Answer 20

The produce of the likelihood of individuals observations in the smaple

Question 21

What is Newton - Raphson method?

Answer 21

Used to find the best model coefficients
- use row^2 to measure the goodness of fit
p^2 = 1 - LL(ß*)/LL(O)

When p^2 = 1 happens in the ideal situation when LL(ß8)= 0

p^2 = 0 model does not tell us anything

ß = maximum likelihood

ln O = likelihood if not preference is given