Professional Documents
Culture Documents
11-12-2019
Eric Molin – TU Delft
Delft
University of
Technology
• previous lecture
• constructing alternatives using basic plans
• this lecture
• constructing choice sets
• sequential and simultaneous construction
• Ngene introduction
1. by basic plans
• published paper schemes
• orthogonal designs only
• easy to use
2. by Ngene
• licensed software
• offers more possibilities (a.o. efficient designs)
• allows evaluating design characteristics
stated choice
Regret experiments
in Traveler Decision Making 6
Constructing choice sets
• choice set
= two or more alternatives to choose from
1. Sequential construction
• construct alternatives
• randomly place alternatives into choice sets
2. Simultaneous construction
• simultaneously construct alternatives and choice sets
stated choice
Regret experiments
in Traveler Decision Making 8
Steps in sequential construction
= unlabeled alternatives
• alternative names do not represent a characteristic
• option 1, option 2; flight 1, flight 2; car1, car2
Imagine this situation applies to Delft station, what would you choose?
stated choice
Regret experiments
in Traveler Decision Making 13
Introduction flight choice case
• Services
• Meal – 2 levels: €0 / €7,50
• Baggage – 2 levels: €0 / €25
• Legroom – 2 levels: 74 / 80 cm
• Number of profiles
• 3 attributes in 4 levels & 3 attributes in 2 levels
• full factorial = 4*4*4*2*2*2 = 43*23 = 512 profiles
• Basic plan 3
• 16 profiles
• Only main effects, no interactions
• two vases:
A. 16 alternatives for alternative A
B. 16 alternatives for alternative B
stated choice
Regret experiments
in Traveler Decision Making 24
Assignment
3. How come?
question 1 question 2
stated choice
Regret experiments
in Traveler Decision Making 35
Bike parking facility at train stations
• Unlabeled experiment
• P – price: 0 / 0,75 / 1,50
• WT – walking time to platform: 0 / 2.5 / 5 min.
• S – surveillance: N(one) / C(amera) / P(ersonnel)
• If choice for Paid in first 3 choice sets, not clear whether prefer
low price, of dislike none surveillance: no info on trade-offs
• Same: choice for Free in last 3 choice sets, not clear whether
prefer Personnel surveillance, or dislike
Regret in high
Travelerprice:
Decision Making 37
37
unlabeled: correlation PA and SB = 1
parking A parking B
PA SA WTA PB SB WTB
0.30 .. .. 0.30 N ..
0.30 0.75 N
0.30 1.50 N
0.75 0.30 C
0.75 0.75 C
0.75 1.50 C
1.50 0.30 P
1.50 0.75 P
1.50 1.50 P
stated choice
Regret experiments
in Traveler Decision Making 39
simultaneous construction of choice sets
stated choice
Regret experiments
in Traveler Decision Making 44
Assignment
• imagine the following stated choice experiment is constructed
• 4 attributes: x1, x2, x3, x4
• each attribute varies in 3 levels
• choice sets of 3 alternatives each: alt1, alt2, alt3
no within-alternative correlations
but between-alternative correlations
• sequential
• no within-alternative ~, but between-alternative correlations
• use for unlabeled alternatives with generic attributes
+ results in a small number of choice sets
• simultaneous
• no within and no between alternative correlations
• use for labeled alternatives with alternative specific
attributes
- typically more choice sets needed
stated choice
Regret experiments
in Traveler Decision Making 50
Assignment
Assignment
• How can Ccar be interpreted?
• Assume that Ccar is positive: can you give reasons for such a finding?
• utility functions:
• Vcar = Ccar + βT * Time + βC * Costs
• Vtrain = βT * Time + βC * Costs
• utility functions:
• Vcar = Ccar + βT_car * Time + βC_car * Costs
• Vtrain = βT_train * Time + βC_train * Costs
• this allows examining whether weighs for time and cost differ
between car and train
stated choice
Regret experiments
in Traveler Decision Making 58
primary question
Which electric car do you prefer?
• disadvantage:
• choice for base alternative provides no information on
trade-offs among attributes
• C = EC constant
• coding of constant: 1 = electric car; 0 = gasoline car
• interpretation
• the base utility difference of EC compared to gasoline car
• is utility when all attributes have value 0
• is determined by all attributes and perceptions associated
with EC that are not variedRegret
in the experiment
in Traveler Decision Making 63
63
Ngene – software for experimental designs
stated choice
Regret experiments
in Traveler Decision Making 64
Ngene
• syntax driven
• utility function is basis
• Vcar = b0 + b1*time
• Vbus = b1*time + b2*price
2. labeled alternatives
• b0 is an alternative specific constant
• Price is only varied for bus
specifies attribute levels for attribute time: hence, 5; 7.5 and 10 minutes
– Ngene automatically replaces design coding with real values
if attribute is generic, you do not need to repeat the attribute levels:
– the same levels are chosenRegret
for inbus andDecision
Traveler for carMaking 69
69
Running Ngene
choice set 1:
1. car
time: 5 min.
2. bus
time: 5 min
price: € 0.75
stated choice
Regret experiments
in Traveler Decision Making 74
estimating interaction-effects
• Vcar = b0 + b1*time
• Vbus = b1*time + b2*price + b3*time*price
• drawback:
• may result in more choice sets
• correlations among interactions may increase
• will usually not cause any problems
Regret in Traveler Decision Making 78
78
blocking
stated choice
Regret experiments
in Traveler Decision Making 79
number of choice sets per respondent
• blocking:
• divide total number of choice sets in smaller blocks
• by free column in the design
thus, blocking
may require
larger designs!
Regret in Traveler Decision Making 83
83
Content computer practical 1
stated choice
Regret experiments
in Traveler Decision Making 85
8
6
Assignment
6. utility functions
• unlabeled: generic parameters only (no constant)
• V1 = βT•Time + βC•Costs + βM•Mode
• V2 = βT•Time + βC•Costs + βM•Mode
stated choice
Regret experiments
in Traveler Decision Making 89
Basic plans 1, 2 & 3
• 11 attributes
• 2 levels each
• masterplan 2
• 7 attributes
• 3 levels each
• basic plan 4
• 13 attributes
• 3 levels each