Overview

The raw airline data below describes the on time and delay departure rates of two airlines, Alaksa and America West, at four airports.

This is an acceptable format for human inspection, but the wide layout and untidy format is not firendly for computational analysis of the airlines’ performance. Namely, mutiple observation of airline performance at all airports are on one row and the statuses, i.e., ON TIME and DELAYED at specific airports are described across two lines. The purpose of this assignment is to restrucutre this data in to a “tidy” format in order to carry out basic analyses of performance.

Relevence

A lot of publicly available data is formatted for human consumption. Wide formats are sometimes easier for the human eye to generalize informaiton, but it is not necessarily the case for computational anlalysis. The ability to transform such data from “wide”" to “long”“, or vice versa, is a necessary skill in analyzing the wide array of data formats available for analyses.

Data Reference and Libraries

The raw airline data is available from the this GitHub repo. Further reading of the underlying problems and solutions for cleaning data is available from Hadley Wickham’s paper, Tidy Data. In addition,

Outside of formatting packages, the libraries used for this project are:

• data.table: Light-weight, fast development for data structure; not neccesarily needed for this data set, but recomended under best practices.
  
• dtplyr: Data table back-end of dplyr for data manipulation.
• tidyr: To produce tidy data from the dplyr pipeline.
• ggplot2: Provides some advanced plotting utilities.

Fill Data

Prior to tidying our data, it would be easier to start off with data with proper column headings and values filled for the NULL observations:

airline_dt$V1 <- rep(airline_dt$V1[!is.na(airline_dt$V1)], each = 2)

colnames(airline_dt)[1:2] <- c("Airline", "Status")

knitr::kable(airline_dt)

Melting and Restructuring Data

Under best practices for a tidy data set, i.e. Hadley Wickam’s three point rule:

• 1. Each variable froms a column.
• 2. Each observation forms a row.
• 3. Each type of observation unit forms a table.

the above data do not conform to the rules. The issues with our airline_dt data mainly vioates rules 1 and 2. Specificaly, rule 1 is violated by the airlines’ status of ON TIME or DELAYED being split over two rows for each respective airline and rule 2 is violated by the multiple departing airports being aggregated under one observation when they should be represented under unique observations. tidyr supplies utilitiies to “melt” the data structure and reform it from its “molten” state. First we will melt the respective airports and their accompanied count’s to be underneath two columns:

airline_dt <- airline_dt %>% gather(Airport, Count, -c(Airline, Status))
airline_dt <- airline_dt %>% arrange(Airline, Airport)

knitr::kable(airline_dt)

In order to conform to rule 2 above, on time and delayed statuses of the airlines should be on under one single observation, being that an airlines performance at a given aiport should be a singular observation.

airline_dt <- airline_dt %>% spread(Status, Count)
colnames(airline_dt)[3:4] <- c("Delayed", "OnTime")

knitr::kable(airline_dt)