Sie sind auf Seite 1von 3

Assignment 1.

Trip Generation Modeling

CIEM 5620 - Traffic Demand Analysis

ASSIGNMENT 1

Question 1. Consider following household data by zone

Zone No Population Households Person Trips


Produced
1 1230 420 517
2 1620 500 621
3 1100 300 408
4 3150 800 945
5 1000 270 422
6 3000 810 1156
7 2175 445 612
8 7000 2210 3074
9 3000 850 1169

where person trips include those made by car and transit. Develop the following work trip
production (WTP) equations, respectively:

WTP  a  b(POP)
WTP  a  b(HHLD)

1. Estimate the parameters in both models


2. Overall model fit (coefficient of determination)
3. Significance of parameter estimates (t test of all b)

Based on the above checks, judging and comparing the quality and acceptability of the two
estimated linear models.

Question 2. Considering the following binary modal-split model between two different zones
(or areas):
exp  t1 
P1 t  
1 1
 
exp  t1   exp  t 2  1  exp  t 2  t1  1  exp  t
exp  t
P2 t   1  P1 
1  exp  t
where t k is the total travel time in travel mode k ( k  1,2 ) and  a parameter to be
estimated.
During the development of a study, travel times were calculated as the average of five
measurements (observations) for both modes, at a cost of $1.0 per observation (one
observation includes observation of both t1 and t2 or $1.0 for one observation of both t1 and
t2 ), and the following values were obtained:
t1  12  2 min , t 2  18  3 min

-1-
Assignment 1. Trip Generation Modeling

(a) If the estimated value for  is 0.1, compute a confidence interval for P1 at a given level
of confidence (e.g. 95%) (suppose a normal distribution for P1 and P2 for simplicity).
(b) Assume you would be prepared to pay $3.0 per each percentage point of reduction in the
error of P1 (Percentage error is defined as the ratio of the standard error to the estimated
(average) value: eP1  P1  100% and eP2  P2  100% ): find out whether in that case it
would be convenient for you to take 10 extra observations of both modes whereby the
following values for t k would be obtained:
t1  12  1min , t 2  17.5  1.5 min
(in your calculation, suppose t1 and t 2 are two independent random variables)

Question 3. Consider the following trip attraction models estimated using a standard
computing package ( t -ratio for each parameter is given in parentheses):

Y  123.2  0.89 X 1 R 2  0.900


(7.3)

Y  40.1  0.14 X 2  0.61X 3  0.25 X 4 R 2  0.925


(1.9) (2.4) (1.8)

Y  1.7  2.57 X 1 1.78 X 4 R 2  0.996


(9.9) (-9.3)

where Y are work trips attracted to the zone, X 1 is total employment in the zone, X 2 is
industrial employment in the zone, X 3 is the commercial employment in the zone and X 4 is
service employment.
Choose the most appropriate model, explaining clearly why (i.e. considering all its pros and
cons).

Question 4. Considering the following two AM peak work trip generation models, estimated
by household linear regression ( t -ratio for each parameter is given in parentheses):

y  0.50  2.0 x1  1.5x2 R 2  0.589


(6.9) (5.6)

y  0.01  2.3x1  1.1Z1  4.1Z 2 R 2  0.601


(4.6) (1.9) (3.4)

where y are household trips to work in the morning peak, x1 is the number of workers in the
household, x2 is the number of cars in the households, Z1 is a dummy variable which takes
the value of 1 if the household has one car and 0 otherwise; Z 2 is a dummy variable which
takes the value of 1 if the household has two or more cars and 0 otherwise.
(a) Chose one of the models explaining clearly the reasoning behind your decision.
(b) Graphically depict both models using appropriate axis.

-2-
Assignment 1. Trip Generation Modeling

(c) If a zone has 100 households (with an average of two workers per household), of which
50% has no cars, 35% has only one car and the rest exactly two cars, estimate the total
number of trips generated by the zone, Oi , with both models. Discuss your results.

----------------------------------------------------------------------------------------------------------------
Hints for Solution of Q2

Part (a)
 Note that t1  12  2 min means that the mean value is t1  12 min and the standard
deviation is et1  2 min .

 Regarding t as a function of  t1 , t2  given by t  t2  t1 , calculate  t (or t ) (mean)


and et (standard deviation)
1
 For a given parameter  , P1  has a single variable t . Regarding P1 as a
1  exp  t 
function of t , calculate  P1 ( P1 ) (mean) and eP1 (standard deviation). The mean value
1
of P1 is simply approximated as  P1  P1 

1  exp t 
 Suppose P1 is normally distributed, then the 1   confidence interval of P1 is given by
P1  P1
 z 2   z 2 . Or alternatively, more precisely, in view of the limited number of
eP1 n


 P1   P1 

samples, n=5, we have P tn 1, 2   tn1, 2   1   .

 eP1 n 

(* Reference: Statistical Analysis for Engineers and Scientists, By J. Wesley Barnes,
McGraw-Hill Book Company, International Edition 1994.)

Part (b)
 When t1  12  1min and t2  17.5  1.5min , using the above method to calculate P1 and
eP1 again.
 Percentage error of P1 is defined as eP1 P1 . Calculate the percentage error in both case (a)
and in case (b), respectively.
 Suppose the percentage error in case (a) and (b) is 14.0% and 8.0%, respectively, then
the number of percentage points of reduction in the error of P1 is 14.08.0 = 6.0 (%)
(Please do your own calculation), and the money prepared to pay for this amount of error
reduction is $3.0  6.0  $18.0 . Because the cost of realizing this error reduction by 10
extra observations is $10.0 < $18.0, we can say that it would be convenient to take the
extra observations.

-3-

Das könnte Ihnen auch gefallen