INTER-CITY MODE CEOICE ANALYSIS USING WLOGIT PACKAGE

A CASE STUDY OP WARANGAL-VIJAYAWADA CORRIDOR

Prof.S. RAGHAVA CHARI

Sr1 C.S.R.K. PRASAD
Sri K.HARIKISHAN REDDY

CENTRE FOR TRANSPORTATION ENGG.

REGIONAL ENGINBERING COLLEGE
WARANGAL-506 004 (A.P.) INDIA.

ABSIRAST developed. This nodel captures the

This study attempts to behavioural aspecte of the choice
variables Lnfluencing the ldent1fy the
mode choice
riderc. Drawing the lnformation from the
behaviour of pa8sengere in model, policy optlons are suggested to
inter-city
travel. A computer package, 'WLOGIT' le obtain
the
the desirable mode
corri or
split along
to avoid underutilisation
developed
(MNL)
based on Multinomial Logit of eystens.
Independence
formulation by b÷passing
of Irrelevant
the
Alternative
(IIA) property. The 'data, obtalned
2.0 METHODOLOGY
through direct interviewing technlque at The widely used MNL model (Moshe E.
selected tovns along
Rail-Road corridor, Warangal-vijayawada
1s used in the
Ben Akiva and Martin G. Richards, 1976,
Chari, 1976: Chari and Khanna, 1978,
analysis. Successive categories method Sinha, 1981, Virendra Kumar and
adopted
attitudinal
in the quant1fication of, 1985, Prasad, 1988, and PonnuswamiKhánna
et al
variables. From the 1992) assumes
analysis, it is found that time, COBt,
a property knovn
comfort and
convenlence variables are Independent rom Irrelevant Alternative
(IIA) implying that the ratio of
significant ir predicting the mode probab1lities of any two modes
choice behavior. However passengers are
observed to be more sensitive to the
independent of other modes. To bypass
this restrictlon,
variation in travel cOst. Suggestions (1979) Gaudry and wA1ls ,
àre made to get the desired Proposed the DOGIT! BOdel
mode share £ormulation through the introduction of
betyeen Ra1l and, Bosd pausenggc.along Theta
the corridor. parameter, which enables to
Incorporate the disutilities of other
modes well, The ratio of
1.0 INTRODUCTION probab111ties of twO alternatives
and 'k ls
There are number Vin
corridorg n transport Vjn
a region n which Bus,
Traln and other services compete Gach (1)
other. The share of travel by Train and
Bus depends on the relative levels of
Pn
service and on commuter behaviour. The wherei
level of service factors lncludes travel
time, travel cost, comfort, convenience, Pin Pn Probabilities of choosing the
reliability and noise
disturbance of lth and kth
modes. i The sensitlvity of people tor
alternatives
from choice set Jn
Vin = Observed
each of the above factors, diflers based
on the affortab1lity and the travel
disutility tunction
of alternative
for an
8ituation. Proper 1dent1fication of
=
individual 'n'.
variables i8 pre-requisite to understand a, *Time + atcost +
and to predíct the mode cholce behaviour acomfort+
value + a, convenience
of passengers. In this paper 8uch as *Reliability value +vlue
as
* Noise
analysis
Vijayawada
1s carried out
Rail-Road
tor
corridor
Warangal
by a,,a. disturbance
disutllity value + b
coefficients
intervi ewing Rail and Bus passengerB to bi = bias variable of i
= Theta weight
assess their attitudes towarda the modes
and level of service offered by them, A
associated with
ith alternative
mode ohoice package 'WLOGIT', by Even ln
Avoiding IIA property of the MNL, the two mode
situation, the choice
'DOGIT' B perfornanco
I|- 33
atleast equal to that of Logit or better It ie assesBed that the
t han that. 1 mportance of
perceived
the dttributes t,hem6eIvpg
Based on this approach, a
are
functjons of characte istics of
computer group of travellers. It ig also
programme WLOGIT' is written in Fortran postulated that the relative
The parameter ai and 1 in rating of
language..
eq.l. are to be obtained by lormlng
modes on the basis of 8ome attribute
a functlon of the real d1£ference that
loglikelihood Eunction and maximising 1t oxista betweer he Tcdes. +he
with reference to those paraneters. critical assumption made is that the
Newton-Raphson algorlthm 1e employed in distr1butlon of response to stimulus
the analysis. The 'WLOGIT' package gives normal on the Psychological continuum
the
coeffleun values of disuti1lty (Guil£ord, l987).
Max. 1ikelihood value ,
statistical summary of
variables, their correlationindependent
matrix, The 8urveys were organised during
the last week of January, 1993, one day
direct and indirect elasticlties .of the
BOdes for various variables. Statistical at each town. They were administered in
two shift8 1.e. 7 A.M. to l P.MI and
tests 11ke Pseudo
erro,
R-square, Standard
T-ratio and percent correctly
P.M. to 7 P.M. simultaneously t both
predicted are also Bus and Railway etationg. In each shift,
"WLOGIT'. incorporated in the 12 trained enumerators (6 at Bus station
and 6 at Railway station) were employed
3.0 The passengers who were waiting to board
RAIL-ROAD PASSENGER STUDIES the BuB or Traln, were randomly selected
and interviewed . The purposewise split
Warangal-vijayawada
corridor is one of the najÍr Rail-Road
Rail-Road of the 762 passengers (both Train and
corridors in the 3tate
Bus) intervlewed lø as follows ;
of Andhra
Pradesh. It connects tive
stations viz. Important Work Business Social Recreation Others
Mahabubabad, Khammam, Kazipet/Warangal,
Madhira and
171 440 311 33
vijayavada. It has a 220 km. of rail The attitudinal data
route and 270 km, road route obtained in
The first stage is quantified using the
ra1lvys operate several express and method of 8uccessive categorles to flnd
pausenger trains on this rail corridor, the scale values, which are given in
wher as the Andhra Pradesh State Road Table 1.
Transport Corporation Operates Super
Delux, Seni-Luxury , Express and Ordinary Table-l: Scale Values
services on the road. of Variables
Thedata required to predict the Varlable Scale value. Ta
vehavior of choice ríders for
inter-city Cost
travel along the corridor is obtained by Time
4.26
direct intervieving 10.59
technique.
fornat is prepared by franing
The Comfort 7.27
related to socio-economic questions
data Reliabillty 6.55
of Conveni ence
riders, level of service varlables of 7.74
Bus andTrain and
Noise disturbance 4.36
mOdes attitudinal
variables 1ike comfort
convenience,
rellab1lity and nolse disturbance.
Previous researchers (Recker and Golob, These scale values are
welghtages for variables inconsidered
as
1976) have Guggested the paychometric disuti1lty
scaling technique as sultable method expresslon. The values of attitudinal
to
quanti£y the attltudi nal variables and varlables are computed as follows ;
hence the s ame ls used to quant1fy them.
The attitudinal variables data Wag Sain (2)
obtained on5-point
a
paychological
continuum in two stages . where
= The quantifled value of the
Stage-l: Relative Lmportance
rating
(Unimportant to extremely important) of attitudinal variable 'a! for
all the varlables. alternative i'for trip maker
'n'.
Stage-2: Relative satisfaction ratings Ia The scale value of attitudinal
(very less to very high) of Traln
and variable 'a,
Bus modes for each attitodínal variable.

I|34
Satn The relative satisfactory level Table 2: Optimal values of Disut1lity
of attitudinal variable 'a'for Coefficients
alternative '1' for trip maker
'n'. Coeflclent Öpt Imum value T-roto
Name
The input data required for model
calibration for each observation ls mode TÍme coell. 0.0885E 4.1
chosen, Travel tlme ( in hrs), Travel Cost coefl. 0.05201 6.4
cost (in Rs.), attitudinal variables (aB Comfort coeff. -0.11786 -7.1
per eq.2) and their weightages. A data Convenlence cocf. -0.05907 -5.5
set consisting of all such observations Bus bias -1.17115 -5.5
in the required format, is propared.
Theta welght for
4.0 MODEL CALIBRATION Traln (e, ) 0.004101
Theta ve1ght for
Initialising the Theta welghts (8 Bus (1) 0.006751
and e)
coefficients
as zeros, the iéutl1Lty
are obtained through Logllkelihood at
WLOGIT' pack age. The coefflcient0 of zero coe flcients -542.4
reliability and 'nolse disturbance' Logl1kelihood at
are found to be Btatintical ly opt ima l values -277.2
insignificant and hence . aro dropped Rho-Index 48.9
from the analysis . Revised scale values
are obtained for the remalnLng
variables. By increment ing Theta díBut ilLty or more wil1 be the utility
Welghts, values of disutility Eor travel. Negative Bus bías represents
coefficients and Log l1kelihood value om1t ted variable
are obtained. This process is repeated
the in
express1on . The results shows that
disutility
until maximum likelihood value 1g
Obtained.
addition to travel tíne and travel cOst.
of
Fig.l. depicts the variation comfort and convenience are also egually
l0çlikelihood value with reference to influencing the node choice behavior of
the Theta weights.
passengers in intercíty travel.
5.0 POLICY SUGGESTIONS
The mode choice
passengers for a given change in
switchingtravel
of
-715 time and travel costs of Bus and Train
modes are carried out. By varying one
variable at a tine, while keeping all
-274
other variables fixed, probabilitles of
Bus and Train nodes are computed. Pig.2
and Fig. 3 shors the probability Curves
for Bus and Train nodes for a variety of
time and cost combinations. The ratlos
-279 in Pigs.2 and 3 rcpresents the catio of
proposed and present values of the
THETA CTHTS
varlable.
100000 +0.00s 00010
00000
fig l LOG LIMUWO00 sUACES
04
The optimum values of Theta welghts
and disut1llty coeflicients at maximum
logl1kelinood are shown In Table 2. It
seen fr om the Table 2, that al1 the
included varlables Are statistical ly
6igni!icant. Time and cost coetticients
Are pos1tive indicating that mOre the
i0rney time, And journey cont Ore
the
diautil1ty for travel
Siml larlji comlort And Convenlence
COef£ iciert negati ve Indicatlng
that mGre t he co fort and Convenence
offered A mde, lesa will be the
 6.0 SUMMARY
)

This paper has presenteu moue

6 choice analysis uf inter ity travel
along Warangal -vi jayawada COrr1dcr.
computer pickage 'WLOGIT' was developed
wIILCi bypaóStS ti.c Irdc;ond1ne
Irrelevant Alternative property
Multinomial Logit Model. From the
analysis of the data collected On
0 25 050 100 150 200 250 300 3 50 40
Warangal-vijayawada corridor, the
TRAVEL TE RATIO FOR TRAN
+10T +12T following conclusions are drawn:
I: TRAYEL COST RATO FOR TRAIN The coefficients of Rellability and

Fg-3 PROBABILITY CURVES FOR TRAIN Noise disturbance are statistically

insigniflcant indicating the<r
The passengers are observed to be linited role in mode choice behavior
Jre
sensitive to the coBt of travel at regional level
lan tine of travel. Hence, the
hows that Btudy attitudinal variables Comfort
ticket Eare the ma jor
The
actor in deciding the mode of travel and convenience are also important
tleast on regional level travel. travel time and travel COst in

The present modal share between

deciding
inter city
the
travel.
mode of travel for
Bus
nd Train along
Jrridor is observed Warangal-vijayawada
to be 25 and 751 The mode choice behaviour
espectJly. Ia order to encourage passengers for variations in travel time
and travel coBt ls 8tudied. The
asseng r to switch from Train to Bug
t is herefore ess ential to passengers are observed to be more
8t of reduce the sensitive for the variation in travel
travel bes ides faster movements
Bus
systems. In additlon to the cost. From this analysis, it 1s ob8erved
DOVe, comfort and convenience levels in that if the Bus fares are reduced by 34*
mode are to be then Bus and Train shares will be equal
improved further.
able 3 shovs the various along the corridor.
et the required Bus share,suggestions to
based on the
nformation gíven in Fig.2. REFERENCES
If the present Bus fares are
68 Or travel times by l5, thenreduced 1. Char1, S.R, 1976. Disaggregated Transnor tat io
the
share will reach 301. Similarly to Demand Mode l for Home-Based Tr ip Gener at ir,
et 50 percent Bus share, the Bus Doctoral Thes1s, Uriver sity of Roorkee, Rccr Pp
tares
re to be reduced by 34. 2. Chari, S.R., Khanna, S.K. 1978. A Stucy cf ork
tr ip trave l in India Jour nal of Transp. Engg.A
able-3: Policy options to get required 3. Gaudr y, M.J.I., Wills, M.J. 1979. The Dogit
Bug share Model Transportat ion Res. Vol. 13B, No.2, pp. 155- 166
4. Gu1lford, J. P, 1987. Psychometr tc Methods, Tata
equired Proposed policy McGraw Hil1 Publish ing Co.Ltd., New De lhi.
Us share 5. Moshe E.Ben Akiva & Mart in G.Richards 1976.
254
Disaggregate Mult1-modal Model for work tr ips
Present condition in the Nether lands, TRR 569,TR8 Washington, pp.
308 61 reduction Ln Bus fares, or 6. Ponnuswam1,S. Anantara jan, T. ,Manju, 107- 123.
151 1992.
reduction in Bus travel time Moda 1 choice in Madras, Ind ian
Highways,
408 211 ceduction in Bus fares, or Vol.20, No.1, pp.27-38.
551 reduction in BuB travel times 7. Prasad, C.S.R.K. 1988. A
50 348 strategy for improving
mass transit patronage for work and other tr ips
reduction in Bus fares
through mode choice ana lys is, M. Tech.Thesis,RECW.
8. Recker, W.W., Go lob, LF. 1976. An
In
general, to enhance the com£ ort moda l choice model, Transpor tat 1on attitudina
Resear ch,
avels in Trains for long Vo1. 10, pp.299-3 10.
assengers and to minímise
journey 9. Sinha, H.K, 1981, ChÍice of Urban
nconvenience due to
the
short
Transportation
Modes fur wor k tr ips Doctorul Thest:,
ravellers, the Bus trip University,Roorkee. RocrkQ
mode may be
ncouraged for short distance l0, Virendra Kumar, Khanna, S.K,
journeys 1985, Validit ion
Yadopting the above policies. of interc ity mode cho ice Disaggregate Mode l,
Ind lan Highways, Vol, 12, No.6, pp.7- 12.