In JJNewkirk/Rfars: Data Tools for Fatality Analysis Reporting System

library('Rfars')
library('dplyr')
library('tidyr')
library('readr')
library('maps')

The Fatal Analysis Reporting System is a nationwide (USA) collection of statistics produced by the National Highway Traffic Safety Administration regarding fatal injuries caused by motor vehicle accidents. The Rfars package was created to make two common data tasks more expedient: reading data and creating tablur and visual summaries. This document will describe general use cases of these two componenets of the package.

Reading Data

One of the primary components to any analysis in R is importing data into an R object. Due to the standard naming convenctions of the file formats in FARS data and the static nature of the fields, the steps to import dat have been built into custom fuctions: fars_read() and fars_read_years()Both functions will read data into R, but with slightly different use cases.

fars_read(filename) will read the file given by filename as if it were FARS data. All columns of the data are preserved and the data is converted to a tibble object with the dplyr package. This class provides a condensed print format, but in most other ways functions identially to data.frames.

In general, it is more efficent to use fars_read() in conjunction withmake_filename(year) which create the standard filename for FARS data when supplied the year of interest. This can then be trustingly passed to fars_read() to gather the desired data, give make_filename(year) doesn't error out of course.

path_to_year<-make_filename(2013)
print(path_to_year)

fars_read(path_to_year)

fars_read_years(year) will read multiple years of data when provided a vector year instead of a single year like fars_read(). fars_read_year() will also call make_filename() automatically for each year, returning warnings for years in which results couldn't be processed. The consequence of this is the requirement that all year files must be in the working directory using the standard FARS filename convention.

The output of fars_read_years() is also distinct from the fars_year() function in that only the year and month columns are returned. Additionally, the output is a list of tiblle objects, not a single tibble. This permits the access of single years of data via list subsetting. Due to the limitation of the output, fars_read_years() will likely be called by a user much less frequently than fars_read() as frequency summaries can be calculated directly with other Rfars functions (as we will see later).

my_years<-c(2013:2015)
fars_data<-fars_read_years(my_years)
print(fars_data[[1]])

Summarizing FARS Data

Rfars has two methods of summarizing data: fars_summary_year() and fars_map_state(). The first function ,fars_summary_year() is a tabular summary of fatal accident frequency. This leverages the fars_read_years() function to read in data for years provided and sumamrizes it in a cross tabulation of the accident counts by year and month.

my_years<-c(2013:2015)
fars_summary<-fars_summarize_years(my_years)
print(fars_data[[1]])

The second function, fars_map_state(stat.num,year), visually summarizes in a plot object summarizing frequency and location by state for a given year. Visualization of freuquency and knolwedge of spatial concentration can provide much more insight to accident prevention and analysis. Maps are drawn from the maps package with the map() function. States must be provided to the function in number form, without leading zeros. The state-state.num pairing can be found in the help documenation for fars_map_state(). Multiple states can be provided to facilitate regional analysis; however, a warning will be issued that the control check for valid state.nums is only done on the first entry. Only one year should be provided.

#lets look at Georgia in 2013
fars_map_state(state.num=13,year=2013)

#let add Alabama and Florida the mix for a view of more of the southeast
fars_map_state(state.num = c(1,12,13),year=2013)