Brian D' Mohr and Jilan Chen

1
SEMCOG Household Travel Survey Data Processing
and Reasonableness Checks

• Brian D. Mohr and Jilan Chen
• Southeast Michigan Council of Governments
• 11th TRB Applications Conference
• Daytona Beach, FL
• May 8, 2007

2
Detroit
3
SEMCOG Region
Communities 234
Population4.9 million
Licensed drivers3.4 million
Annual VMT49 billion
Miles of road23,000
4
Presentation Topics

• Summary of data processing and reasonableness
  checks performed on 2004 household travel survey
  data
• Survey background information
• Calculation of survey expansion factors
• Future initiatives and lessons learned

5
Why Collect New Household Survey Data in 2004?

• New snapshot of regional travel behavior needed
• Previous survey conducted in 1994
• Shorter term enhancements planned for four-step
  model
• Possible future move to activity-based model
• Opportunity to partner with MDOT

6
2004 Household Travel Survey Background

• Combination of two household surveys
• Michigan Travel Counts
• SEMCOG Travel Counts
• Survey similarities
• Consultants (MORPACE, PB, Brogan)
• Activity-based survey design
• Survey methodology
• Relational database structure

7
(No Transcript)
8
QA/QC Measures During Data Collection

• Computer-Assisted Telephone Interviewing (CATI)
  logic checks
• MORPACE post-processing checks
• Parsons Brinckerhoff interim audits
• SEMCOG interim audits
• Review of questionable records
• Number of persons, workers, autos per household
• Distributions of trip rates and trip lengths

9
SEMCOGs Post-Processing Data Checks
Database Integrity Checks
Individual Field Checks
Intra-Record Checks
Inter-Record Checks
Distribution Plots
10
Database Integrity Checks

• Checked primary keys for each table
• Checked relationships among tables
• Person ? household
• Household ? person
• Trip ? person

11
Individual Field Checks

• Determined if attribute values fell within valid
  ranges
• Corrected obvious errors
• Found explanations for unusual errors, clarified
  confusing field definitions

12
Intra-Record Checks

• Date versus day of week
• Related age fields
• Related transit pass/cost fields
• Related school variables, work variables
• Fields containing geocoding information
• Origin/destination, arrival/departure fields
• Trip-table fields related to travel modes, travel
  costs, number of passengers

13
Inter-Record Checks

• Arrival location, time compared to subsequent
  departure location, time
• Destination activity compared to subsequent
  origin activity
• Trip characteristics for members of same household

14
Distribution Plots

• Distributions plotted for travel times,
  distances, speeds, activities
• Distributions stratified by mode, purpose,
  geographic area
• Useful for identifying outlying data

15
Assessment of Data Quality

• Overall assessment
• Excellent data quality
• Vast majority of checks uncovered no errors
• Specific findings database integrity, individual
  field checks
• Trip records discovered for immobile
  participants
• Definition clarified for stop field

16
Assessment of Data Quality

• Specific findings (intra-record, inter-record,
  distribution checks)
• 22 records with incorrect day of week
• 587 locations missing geocoding attributes
• 29 records with identical arrival time and
  subsequent departure time
• Work trips found for households with no workers
• Outliers found in some distribution plots

17
Household Geocoding Checks

• All household locations mapped for both MDOT,
  SEMCOG surveys
• Used to separate households in region from
  households outside of region
• Used to check county attribute values

18
Consultation with Parsons

• Suggestions for performing specific data checks
• Opinion on reasonableness of basic survey
  statistics
• Assistance on combining two surveys
• Assistance with calculating expansion factors

19
Combining the Surveys

• Concerns with second day of MDOT survey
• Personal trip-rates dropped from 3.64 to 3.19
• Zero-trip households increased from 8.1 to
  11.0
• Decisions
• Combine only first day of MDOT survey with SEMCOG
  survey
• Calculate, apply expansion factors after
  combining surveys

20
Survey Expansion Issues

• Household size, auto ownership, number of workers
  64 stratification cells
• Spatial stratification (preferably by county)
• Lack of sufficient samples in some cells
• Balancing desire for precision, need for
  aggregation

21
Calculating Expansion Factors

• Cells with insufficient samples aggregated
• Initial expansion factors proposed based on
  experience from other urban areas
• Four-dimensional algorithm by Parsons used to
  calculate final expansion factors

22
Using Draft Expansion Factors
23
Using Final Expansion Factors
24
Future Initiatives

• Perform additional QA/QC checks
• Analyze transit-focused survey dataset
• Develop detailed survey analysis report
  (including 1994/2004 data comparison)
• Develop summary report (regional snapshot for
  public/media)
• Use data in model

25
Lessons Learned

• QA/QC essential from data collection through
  post-processing
• One travel day sufficient for our needs
• GIS useful tool for performing checks
• Four-dimensional expansion factor calculation
  possible

26
SEMCOG Household Travel Survey Data Processing
and Reasonableness Checks

• Brian D. Mohr and Jilan Chen
• Southeast Michigan Council of Governments
• 11th TRB Applications Conference
• Daytona Beach, FL
• May 8, 2007