STATISTICAL EVALUATION OF TRANSIT ONBOARD SURVEY RESULTS AND METHODS Kathleen Yu, Arash Mirzaei,...

STATISTICAL EVALUATION OF STATISTICAL EVALUATION OF TRANSIT ONBOARD SURVEY TRANSIT ONBOARD SURVEY

RESULTS AND METHODSRESULTS AND METHODS

Kathleen Yu, Arash Mirzaei, Behruz Paschai, Hua Yang

North Central Texas Council of Governments (NCTCOG)

FTA Workshop on Travel Forecasting for New Starts 2March 2009

Analyzing the 2007 DART On-Board Survey

Dallas Area Rapid Transit (DART) Transit System Information

On-Board Survey, Spring 2007

Transit ModeWeekday

BoardingsNo. of Routes

Commuter Rail 9,000 1

Light Rail 62,000 2

Express Bus 6,700 10

Local Bus 141,000 98

Total DART System 218,700 111

May 2009 3TRB National Transportation Planning Applications Conference

DART System Map

DART Onboard Survey Schedule

December 2006 RFP posted.

February 2007 Proposals reviewed.

March 2007 Contract signed.

Pre-tests almost eliminated due to time.

May 2007 Data collection ended.

November 2007 Final product delivered

March 2008 NCTCOG clean-up completed.

Weekday Sampling Plan Sample size based on daily route boardings.

Desired: 95% confidence interval with +/- 5% error Random sampling of vehicles by route direction in four

time periods (AM, mid-day, PM, and evening)

For each sampled vehicle: Total adult (15+) boarding counts by stop All adults boarding bus asked to participate

Expansion Plan ExampleSample Vehicle Trip in Route = 45, Time Period = a.m., Direction = Northbound, Day = Weekday

BusStop 1

BusStop 2

BusStop 3

BusStop 4

# of Boardings 15 8 27 10

# of Completes 6 0 9 2

Stop Response Fact. 2.5 0 3.0 5.0

Calculated Boards 15 0 27 10

Route Trip Factor = 60/52 = 1.151.15 1.15 1.15 1.15

+ + + = 60

+ + + = 52

Trip Factor

Vehicle Trip Expansion Factor = Total Vehicle Trips per RTDD / Sampled Vehicle Trips per RTDD = 15/2 = 7.5

7.5 7.5 7.5 7.5Vehicle Trip Exp.

21.6 0 25.9 43.1RTDD Factor 90.6+ + + =

Total Calculated Boardings in all RTDs = 582.4 Observed Avg. Daily Boards = 971 971/582.4=1.7

1.7 1.7 1.7 1.7Route Factor

36.7 0 44.0 73.3Final Exp. Factor8

Quality Control Checks Evaluation of geo-coding

Origin and destination Boarding location (recorded) Alighting location (imputed)

Correction of conflicting answers Identification of inconsistent answers Routes in the paths Boarding and alighting locations

Assessment of the quality of final database

Evaluation of Geo-coding A reasonable qualitative question is: how

accurately were origin and destination locations coded in the final database?

Practically, we accepted some level of inaccuracy in locating the true locations and determined how often the coding has been done within that distance.

Instead of manually checking all 6,447 records, we randomly selected records to assert an accuracy statement based on statistical sampling.

Evaluation of Geo-coding – Results Origin point is within 0.75 miles of the user-

specified origin place and address. 71 of 74 random records checked met criteria. 95% confident that 90 to 100% of all points in database

meet criteria.

Destination point is within 0.75 miles of the user-specified destination place and address. 72 of 74 random records checked met criteria. 95% confident that 90 to 100% of all points in database

meet criteria.

Questions1. Home address2. Trip origin address and type3. Mode of access4. Transfer from5. Total number of transit vehicles

in the trip6. Transfer to7. Mode of egress8. Trip destination address and

type9. First and last rail lines and

stations boarded and alighted 10. Route sequence11. Main reason for taking the

route12. Perceived trip length in minutes

13. Frequency of use14. Weekend and/or weekday users15. Substitute mode of travel16. Type of payment for the transit17. Type of fare18. Number of registered vehicles19. Household size20. Number of adults (15+) in the

household21. Male or female22. Age group23. Work status24. Ethnicity25. Household income group

Identification of Inconsistent Answers 4 to 6 redundant questions about the path possibly

confused people

2,593 of 6,447 records found to have inconsistent path information.

Path Response Errors People described all possible routes they could

take for their trip and not the ones they are specifically taking on this trip.

People described a round trip and not a one-way trip, so a route/rail was repeated in the sequence.

People put down origin and destination for round trip, but described path for one-way trip – or vice versa.

People described their reverse trip rather than what they were taking. Additional 575 surveys were appended to survey review

list because the origin seemed closer to last route than first route in sequence.

Correction of Paths1. Create a map with the origin, destination and all routes

reported in the answers to the redundant questions.

2. Assume origin, destination, and surveyed “route” are correct.

3. Review routes in path sequence and the routes in the transfer questions to see if they contain a reasonable path.

4. If needed, use any routes listed on the survey to create a reasonable path for the OD using the surveyed “route.”

5. For walk mode of access and egress, confirm distance is less than or equal to 2.5 miles.

Of 3,168* surveys reviewed, 3,004 surveys were checked, modified (if necessary), and returned to the database, and 164 flagged for uncorrectable paths.

* Original list of 2,593 records plus the 575 surveys with possible reverse path sequence.

Examination of Path Questions

Path Transfer Questions Sequence Question

No Transfer 86% 98%

1 Transfer 73% 73%

2 Transfers

3 Transfers 24%

Overall 60% 72%

Percentage of Correct Answers to Path Questions

Confidence in Path Sequence Random selection of 74 records from the database

of 6,283 records. 71 determined to have a correct path. Assertion: 95% confident that 95% of the respondent-

identified sequences from origin to destination is feasible.

Potential Survey Improvements Rigorous sample expansion plan.

Detect and correct for non-response biases Ancillary count program

Pilot tests conducted in a timely manner. Test graphic design Clarify meaning of one-way trip

Reduction in number of questions. Elimination of redundant questions. Alighting location asked or captured. Transit wait time asked or captured. Clarification of expected quality in surveyor contract.

FTA Workshop on Travel Forecasting for New Starts 19March 2009

Planning for 2008 On-Board Surveys in Fort Worth and Denton

FWTA and DCTA Transit System Information

Transit ModeWeekday

BoardingsNo. of Routes

Express Bus 1,126 9

Local Bus 32,882 53

Total FWTA and DCTA 34,118 62

On-Board Survey, Fall 2008

FWTA = Fort Worth Transportation Authority

DCTA = Denton County Transportation Authority

FWTA and DCTA System Map

FWTA and DCTAOnboard Survey Schedule

June 2008 RFP posted.

August 2008 Proposals reviewed.

Contract signed.

October 2008 Pilot test/cognitive interviews conducted.

November 2008 Data collection ended.

January 2009 Draft geo-coded database delivered.

April 2009 Expanded database delivered.

June 2009 Final report scheduled.

Improvements Implemented

Clarification of expected quality in the survey contract Reduced number of questions

Unused questions in DART 2007 survey eliminated Redundant path questions in DART survey removed

Inclusion of new questions Alighting location asked or captured Transit wait time asked or captured

Clarification of meaning of one-way trip Use of graphic diagrams Advertisement campaigns Consistent use of term “one-way trip”

Improvements Implemented

Enhanced questionnaire design Modified text design Graphic design

Pilot tests of different instruments In-field Cognitive interview

Sample expansion plan Sampling expansion correction by station counts Non-response follow-up survey

Clarification of Expected Quality in Surveyor Contract Identify the sample frame, confidence interval, and error

Daily boardings per route, 95%, and +/- 5%.

Identify QC specification.The consultants should show they are 95% confident that 90 to 100% of location data and all other collected information in the survey instrument are coded correctly. This quality will be tested by random sampling of the final coded records. A record is considered incomplete if ….

Identify QC implementation plan.At the end of the data collection, the consultants should show that they are 95% confident that 90 to 100% of the collected surveys are useable.

A filled questionnaire is considered useable if the respondent properly answered all of the following questions: origin, destination, purposes, path questions that identify the routes that the users take, mode of access and egress, time of survey, and home address.

Unused Questions in DART 2007 Survey1. Home address2. Trip origin address and type3. Mode of access4. Transfer from5. Total number of transit

vehicles in the trip6. Transfer To7. Mode of egress8. Trip destination address and

route12. Perceived trip length in

minutes

13. Frequency of use14. Weekend and/or weekday users15. Substitute mode of travel16. Type of payment for the transit17. Type of fare18. Number of registered cars19. Household size20. Number of adults (15+) in the

Redundant Questions in DART 2007 Survey1. Home address2. Trip origin address and type3. Mode of access4. Transfer from5. Total number of transit

vehicles in the trip6. Transfer to7. Mode of egress8. Trip destination address and

route12. Perceived trip length in

minutes

13. Frequency of use14. Weekend and/or weekday users15. Substitute mode of travel16. Type of payment for the transit17. Type of fare18. Number of registered cars19. Household size20. Number of adults (15+) in the

Clarification of One-Way Trip Use of graphic diagrams. Advertisement campaigns:

Rail hangers in Fort Worth Posters in Denton

Consistent use of the term “one-way trip.” Consistent font throughout the questionnaire.

Use of Graphic Diagrams

Advertisement Campaign – FWTA

Advertisement Campaign – DCTA

Consistent Use of Term “One-Way Trip”

Questionnaire Design Modified text design:

Graphic design:

Modified Text Design

Graphic Design

Pilot Tests In-field: test on buses of two questionnaire designs

(NCTCOG staff). Cognitive interview (consultant staff)

People recruited and interviewed for one hour. Asked to fill out both questionnaire designs. Discussion on how they answered each question and

what they did like and did not like in each questionnaire.

In-Field Pilot Test Observations Shaky buses made writing hard. Early morning dark conditions in the bus should be

considered in design. Font sizes should correspond to all users (some

users may not use their glasses in the bus). The placement of the Advertisement poster on a

clear glass made it difficult to read. Informal interviews conducted with respondents to

learn more about riders’ view of the transit system.

By personally conducting the pilot test, NCTCOG staff got a realistic feeling of what surveyors and respondents experienced.

Pilot Test Results Comparative in-field test.

Modified text: 69% completion rate Graphic: 50% completion rate

Cognitive interviews. 28 interviewees Mixed response to both questionnaires.

Pilot Test Conclusions Modified text design was marginally more

successful than graphic design. Tests of graphic design were not comprehensive enough

to reach a solid conclusion. Final design incorporated a few elements of the graphic

design.

In-field pilot test was more useful. Cognitive interview method might be useful when

there are many unknowns about which to decide.

Final Questionnaire Design

Sample Expansion Plan Boarding counts by stop

APC counts not available. Manual counts for surveyed vehicles only.

Station counts by mode of access/egress Major park-and-ride and transfer stations.

Non-response follow-up survey Personal interviews of people not returning the

questionnaire. Measurement of bias in the filled surveys.

Non-Response Follow-up Personal Interview Form

NCTCOG (The T) On-Board Non-Response Survey

Assignment #: ____________ Trip #:______________ Route #:________________________

1) Reason for not taking/completing a survey: Never participate Too many questions Not interested/Don’t Care Conditions on bus not suitableNo time to complete it on this trip / trip too short Other (specify): _______________________________

2) How many minutes will you be traveling on THIS BUS for THIS TRIP? 5 or less 6-10 11-15 16-20 21-25 26-30 More than 30

3) Age 15 - 24 25 - 34 35 - 44 44 - 54 55-64 65+

4) Ethnicity White African American Hispanic Asian Native American Other (specify): ______________________

5) HH Income 2007 <10K 10K–14.9K 15K–24.9K 25K–34.9K 35K-49.9K 50K–75.9K 75K+

Non-Response Survey Results

Survey Question # Responded % Response

Reason for refusal 1,424 98.27%

Minutes traveled 903 62.32%

Age 1,332 91.93%

Ethnicity 1,366 94.27%

Income 500 34.51%

Response Rate to Questions

Total Participation: 1,449 surveys (Early estimate of 87.8% response rate)

Acknowledgment FTA staff: for providing ideas and help in analyzing

the results: Jim Ryan Ken Cervenka

NCTCOG Travel Model Development staff: for managing the project, analysis, and presentation: Arash Mirzaei Behruz Paschai Hua Yang

STATISTICAL EVALUATION OF TRANSIT ONBOARD SURVEY RESULTS AND METHODS Kathleen Yu, Arash Mirzaei,...

Documents

Transcript of STATISTICAL EVALUATION OF TRANSIT ONBOARD SURVEY RESULTS AND METHODS Kathleen Yu, Arash Mirzaei,...

Robust Optimal Adaptive Trajectory Tracking Control of ... · M. Navabi and H. Mirzaei / Robust Optimal Adaptive Trajectory Tracking Control of Quadrotor Helicopter 1041 Latin American

Unido.org/statistics Analysis of Divergence of quarterly and Annual Index of Industrial Production Shyam Upadhyaya, Shohreh Mirzaei Yeganeh United Nations.

Artificial Immune Systems Razieh Khamseh-Ashari Department of Electrical and Computer Eng Isfahan University of Technology Supervisor: Dr. Abdolreza Mirzaei.

Review Article The Possible Role of Novel Coronavirus 2019 ...Hadi Esmaeili Gouvarchin Ghaleh1, Mohammad Reza Karimi 2, Parisa Rezayat , Masomeh Bolandian1, Majid Mirzaei Nodoushan1,

o g r ap h y n Mirzaei and Karimi ceanogr Mar Res 21 5:1 a n M … · 2019-02-15 · ao Mirzaei M, Karimi Z (2017) Modeling of Tidal Currents in the Western North of the Persian Gulf.

BEHRUZ MOFIDI SHIRAZI 90 01 31 0233 TWO, THREE, FOUR, …...behruz mofidi shirazi page 1 90 01 31 0233 two, three, four, five, four, three, two, one. this is detective skuta. today's

SOCIAL LEARNING STRATEGIES IN EFL CONTEXTrjelal.com/3.4.15/63-74 ARSHAD MIRZAEI KHOSHALANI.pdf · Piaget, the main advocate of cognitive constructivism, assumes learning as an actively

Optimization Methods in a Stochastic Production Environmentpure.au.dk/portal/files/100477051/Reza_Pourmoayed_PhD... · 2016-06-07 · Marand, Samira Mirzaei, Maryam Ghoreishi, Parisa

AndrODet: An adaptive ...omid-mirzaei/files/J5_0aucb77k.pdfFutureGenerationComputerSystems90(2019)240–261 Contents lists available atScienceDirect

FAZEL MIRZAEI : AUTOMATED BRAIN TUMOR SEGMENTATION …ictactjournals.in/paper/IJIVP_Vol9_Iss1_Paper8_1844_1848.pdf · brain tumor segmentation are carried out mostly by using MR images

Fine-grained acceleration control for autonomous ... · Fine-grained acceleration control for autonomous intersection management using deep reinforcement learning Hamid Mirzaei Dept.

Benefits and challenges of OER By Farshid Mirzaei

Increasing Precision in Highway Volume through Adjustment of Stopping Criteria in Traffic Assignment and Number of Feedbacks Behruz Paschai, Kathy Yu,

Gaussian Based Hue Descriptors - Computing Science ...funt/Mirzaei+Funt_GaussianHue_PAMI2015.pdfpseudo-color atlas as the representation of hue. Such Gaussian-like reflectances have

Unido.org/statistics Seasonal Adjustment of National Index Data at International Level Shyam Upadhyaya, Shohreh Mirzaei Yeganeh United Nations Industrial.

Mirzaei-Paiaman and Masihi 2014

Mirzaei-FractureMechanics Lecture.pdf

Establishment of Freeway Link Volume Validation Targets based on Traffic Count Distributions in the Dallas-Fort Worth Region Behruz Paschai, Arash Mirzaei,

Rahim Mirzaei : Assistant professor in Institute of Scientific and Applied Higher Education of Jahad e Agriculture.

Evaluating GPS Technology Used for Household Surveys Kathy Yu, Arash Mirzaei, Behruz Paschai North Central Texas Council of Governments (NCTCOG) 15 th.