Data Analysis From Start to Finish

ID 529: Data Management and Analytic Workflows in R

Dean Marengi | Wednesday, January 17^th, 2024

Context for what we’ll be discussing

• To walk through the methods implemented with R to go from raw data to data analysis for an exposure assessment project for “EH 263 Analytical Methods and Exposure Assessment”.

• The project aimed to explore the additional contribution of food truck exhaust to Ultra-fine Particle Concentrations in ambient air, above and beyond background levels.

Methods implemented that we have discussed

• Project based workflows
• Reading in data with readr
• Conditionals and for loops
• Writing functions
• Data manipulation using dplyr
• Combining (linking) datasets
• Text processing using stringr
• Working with factors, dates, and times using lubridate
• Data visualizations using ggplot
• Functional programming for efficient file imports and modeling

Why food trucks?

• UFPs have an aerodynamic diameter of 0.1 µm or less (Li et al., 2016; Moreno-Ríos et al., 2022)

• Diesel exhaust substantially contributes to UFP concentrations

• Food trucks often use diesel or gas generators to power truck operations
  • These generators are typically run for many hours (i.e., for the duration of time trucks are on-site)

Existing regulations

• To improve the ambient air quality of air pollutants from idling vehicles, the Massachusetts Anti-Idling Law limits engine idling up to 5 minutes (MA Department of Environmental Protection, n.d.).
  • However, there are no restrictions in the Mass. Anti-Idling Law, or other laws, that aim to constrain air-pollution emissions from food truck generators)

Hypotheses

1. UFP concentrations will be higher when food trucks are present compared with when they are not present

2. UFP measurements at a 5-meter distance from food trucks will be associated with higher UFP concentrations when compared with measurements taken at a 10-meter distance

Study design overview

Figure A: Sampling site configuration at the HSCP. Condensation Particle Counter (CPC) placement in relation to the HSCP outdoor dining area and food truck parking area is shown in red.

Figure B: Approximate timeline for the sampling protocol. Data collection began at approximately 10:00 AM and ended at approximately 11:30 AM. Food truck arrival times varied, but generally occurred between 10:15 and 11:00 AM.

Data processing problems to solve

Part 1: Compiling and organizing UFP data

• Compile 34 raw data files that were exported from CPC device software

• For each file (Part A):
  • Standardize the file structure
  • Derive new variables based on existing columns, and metadata embedded in file names
    • A sample ID variable
    • File name
    • Several date/time variables
    • An indicator for which CPC device collected the measurements
    • An indicator for which distance the measurements were collected at (5- versus 10-meter)
  • Re-organize the columns

CPC raw data example file (n=34, each with 5000+ rows)

Part 1: Compiling and organizing UFP data (cont.)

• For each file (Part B):
  • Reference a sample log to “look up” truck arrival times for a given sample ID (e.g., FT001, FT002, etc.)
  • Use this truck arrival time to create an indicator for whether a one-second measurement was taken before or after truck arrival (i.e., Pre-truck versus Post-truck)
    • Note: This is because the CPCs were continuously collecting measurements for the whole measurement period (all pre- and post-truck arrival measurements were contained within a single file)

Sample log example file (n=1)

Part 2: Joining measurement and sample log data with covariate data

• The Pre- versus Post-truck indicator in Part 1 allows us to join data from a third file containing covariate data

• The covariate data file contained:
  • Sample ID
  • A pre- versus post-truck indicator variable
  • Meteorological parameters (e.g., temperature (F), % relative humidity, wind speed (mph), etc.

Covariate data example file (n=1)