Tasks

Travel model data prep

Basic processing

• Input: raw survey household, person, and trip files
• Output: household, person, and trip files formatted to model data specifications and filtered/modified to match the model trip domain.
• Subtasks:
  • Spatial join to model TAZ systems
  • Recode raw survey field names and data values to match the travel model specifications
  • Drop out-of-region travel
  • Handling night spent at non-home locations
• SFCTA scripts:
  • spatial join
  • survey recode

Imputation

• Input: Dataset with no imputed variables
• Output: New columns for imputed variables; imputation quality report for each variable imputed
• Transparency: algorithm/model used for imputation documented and reproducible; imputed variables documented; imputation confidence variable? (There appears to be a fair number of papers on how to evaluate imputation methods; we might settle on some. What does RSG use?)
• MTC: RSG 2020 work:
  • https://github.com/BayAreaMetro/tnc_california_studies/tree/stable/8.4_mode_imputation
  • https://github.com/BayAreaMetro/tnc_california_studies/blob/stable/8.3_num_travelers/01_update_num_travelers.R

Trip linking

• Input: Unlinked trips, rules for determining linked trip attribute from unlinked trip attributes
• Output: Linked trip table with linked trip attributes
• Transparency: algorithm used for trip linking documented and reproducible
• MTC: RSG 2020 work: https://github.com/BayAreaMetro/tnc_california_studies/tree/stable/8.2_transit_improvement
• SFCTA: trip linking

Tour building

• Input: Household tables, person tables, linked trip tables; rules for determining tour attributes from linked trip attributes
• Output: Tour table with tour attributes
• Transparency: algorithm used for tour building documented and reproducible
• MTC: RSG 2020 work: ???
• SFCTA: https://github.com/sfcta/travel_diary_survey_scripts/tree/master/2022/03-tour_extract

Weighting

• Input: Household tables, person tables, tour tables, linked trip tables; linked trips/tours have day of week associated with them; requirements for weighting properties
• Output: Input tables with weighting columns, with:
  • Dummy rows for households/persons/tours/linked trips not surveyed
  • Weight columns
  • Report on how weight columns matched the requirements
• MTC: RSG 2020 work: https://github.com/BayAreaMetro/tnc_california_studies/tree/stable/8.1_PopSim_weighting