Skip to contents

Retrieve and summarize sss data

retrieve-and-summarize-sss-data.Rmd

For use with the 2025 PSRC Safety & Security Survey

The Safety & Security Survey is a more traditional attitudinal survey, with responses at a person level, rather than multiple related tables. (The travelSurveyTools package is not used in handling it.) Use get_psrc_sss() to retrieve the variables you need, then use psrc_sss_stat() to calculate weighted counts, shares, or numeric summaries.

Because the package queries Elmer, you’ll need to be connected in office or through VPN to run the retrieval step.

Data retrieval

Since psrc.travelsurvey uses Elmer, the agency’s central database, you’ll need to be connected in office or through VPN. Then use get_psrc_sss(), which takes a single argument:

  • survey_vars - a vector of desired survey variable names

The return object is a person-level data.table. person_id, hh_id, sample_segment, and person_weight are included automatically, so they do not need to be added to survey_vars.

library(psrc.travelsurvey)
library(magrittr)
library(dplyr)

vars <- c("employment", "crash_participant", "travel_anxiety", "safety_choices", "crash_number_people")

sss_data <- get_psrc_sss(survey_vars = vars)

Summarization

psrc_sss_stat() assumes a person-level analysis unit. The main arguments are:

  • sss_data - the table returned by get_psrc_sss()
  • group_vars - one or more grouping variables, in nesting order
  • stat_var (optional) - a numeric variable for min, max, median, and mean summaries

Like psrc_hts_stat(), it uses the last grouping variable as the share variable when stat_var is omitted.

Count and share example

This example estimates the weighted distribution of travel_anxiety responses within each crash_participant category.

rs1 <- psrc_sss_stat(
  sss_data,
  group_vars = c("crash_participant", "travel_anxiety"),
  incl_na = FALSE
)

head(rs1)

The resulting table can be read as an association summary: within each crash_participant group, compare the weighted shares across travel_anxiety response levels.

Numeric summary example

TSS variables are primarily dichotomous or ordinal, but psrc_sss_stat() can also calculate handle weighted means for any numeric variables, using the stat_var argument.

sss_data <- mutate(sss_data, crash_number_people_numeric = if_else(
    is.na(crash_number_people), NA_integer_, as.integer(stringr::str_extract(crash_number_people, "^\\d+"))))

rs2 <- psrc_sss_stat(
  sss_data,
  group_vars = "crash_participant",
  stat_var = "crash_number_people_numeric",
  incl_na = FALSE
)

head(rs2)

Notes

Margin-of-error columns are returned where the underlying survey calculation produces standard errors; if a statistic cannot support a variance estimate, its MOE field will be NA.