Travel Behavior Inventory Survey Helper

This repository contains a collection of R code for working with Metropolitan Council Travel Behavior Inventory Household Survey data.

If you came for a pretty R Data object of the TBI survey data, check out the data folder or source from URL:

load(url(paste0(
  "https://github.com/Metropolitan-Council/metc.tbi.helper/raw/main/data/",
  "tbi.rda"
)))

Other things you might find useful in here include R scripts to create complex cross-tabs that require funky joins, some scripts for appending auxiliary datasets & geographic information to the datasets, and more.

This repository is in active development, and anyone can contribute 🤝. See the CONTRIBUTING page on the right to learn more.

Organization: what’s here

All TBI survey years (2019, 2021, and 2023) are combined in the same object, distinguished by the survey_year column.

• metadata: Read the documentation 💘
• data: compiled datasets generated from raw TBI data, that live in the Council’s database.
  • tbi.rda is a compressed list object containing:
    • dictionary of variable names, values, survey questions, and logic;
    • person-level records (person);
    • day records, for analyzing daily trends;
    • trip table;
    • household (household) records;
    • vehicle (vehicle) records, including fuel efficiency data; and
    • trip_purpose, for working with trip purpose data. This table has been specially weighted to attribute weights to either end of non-home-based trips, and to the non-home based end of home-based trips;
    • mode_conflation.

tbi.rda is the only data we store in this GitHub repository. Git (even Git LFS) is not ideal for storing data. If you generate additional datasets in your work, please add them to the .gitignore file. If you need to work with .csv data, please see the script data-raw/99-distribute-data.R, and add the .csv’s to your gitignore file. If incorporating a new dataset, write it to the database first (see @eroten for write access).

• data-raw: scripts to generate datasets. Work here if you want to add a new variable to the dataset(s) or incorporate a new dataset to the database and/or .RData object. You will need access to internal databases for this work and have certain environment variables in your .Renviron. Contact a project manager (@eroten, @Brandon-Whited) for assistance.
  • data-raw/_data-compile.R is the main script that sources all numbered .R scripts in this folder.
  • derive-var-[variable-name].R: derive new variables from TBI datasets.
  • get-[dataset-name].R: incorporate auxiliary datasets (e.g., fuel efficiency data). If external to the Council, they are added to the database. Wherever possible, spatial datasets are loaded from the Council’s internal GISLibrary.
• R: collection R code to access and perform analyses.
  • R\previous_code deprecated code best used as a starting point for new analyses.
    • xtab-[unit of measurement]-[variable1]-by-[variable2].R: custom crosstabs.
    • fun-[does-this].R: functions or processes.
    • fig-[description].R: create a figure or plot
    • table-[description].R: create a table
    • var-[description].R: create a new variable
    • trim-[description].R: reduce/filter the dataset based on a variable.
• Note that there are no folders for specific TBI-related projects/analyses. These should live in their own GitHub repositories, and source code from this repository.

Creating a custom crosstab

We generally use the tidyverse-friendly package srvyr to create weighted summaries and crosstabs from the TBI data. A simple example, calculating average vehicle age by household income, follows:

library(dplyr)
library(purrr)
library(bit64) # for looking at big integers, like the person_ids
library(srvyr)

# load dataset
load(url(paste0(
  "https://github.com/Metropolitan-Council/metc.tbi.helper/raw/main/data/",
  "tbi.rda"
)))

# create 2021 only dataset
tbi21 <- purrr::map(tbi,
                    filter,
                    survey_year == 2021)

tbi21$vehicle %>%
  select(hh_id, veh_age) %>% 
  left_join(tbi21$hh) %>%
  as_survey_design(weights = "hh_weight") %>%
  group_by(income_broad) %>%
  summarize(veh_age_avg = survey_mean(veh_age, na.rm = T))