This assignment will examine arrival-delay performance for two airlines across five destinations. The goal is to move from the original wide-format table to a tidy dataset that can be used to compare airline performance both overall and within each destination. An important part of the analysis will be determining whether the overall comparison tells the same story as the city-by-city comparisons.
I will first recreate the provided airline-delay table in a CSV file using the same general wide structure as the source. I will preserve the empty cells from the original data as missing values rather than replacing them while creating the file. The CSV will then be placed in a public GitHub repository so the final R Markdown analysis can read the data from an internet-accessible source and remain reproducible.
After importing the data into R, I will inspect the structure, column
names, data types, and missing values before making transformations. I
plan to use tidyr and dplyr to reshape the
data from wide to long format so that airline, destination, arrival
status, and flight count can be represented as separate variables.
Missing observations will be handled explicitly in the analysis rather
than being silently ignored.
Once the data is tidy, I will first examine counts to confirm that the transformed data still represents the information in the original table. I will then calculate percentages so that the two airlines can be compared fairly even if they have different numbers of flights. The analysis will compare the percentage of delayed or on-time arrivals for the two airlines overall and then repeat the comparison separately for each of the five destinations. Tables and/or visualizations will be used to make these comparisons clear.
Finally, I will compare the overall results with the destination-level results. If the overall airline comparison differs from the city-by-city comparisons, I will describe the discrepancy and investigate how differences in the number or distribution of flights across destinations may contribute to it. The conclusion will summarize what the percentages show about the two airlines and explain why examining both aggregated and destination-level results is important.
One challenge will be preserving the structure of the original table while still representing its empty cells correctly as missing data. I will need to distinguish a truly missing value from a count of zero because those values have different meanings. I will also need to verify that the wide-to-long transformation does not duplicate, drop, or mislabel any observations.
Another challenge is making a fair comparison between the airlines. Raw delay counts alone may be misleading if one airline has more flights overall or serves a different number of flights to each destination. For that reason, the main comparisons will use percentages rather than counts alone. I will also check totals before and after tidying to make sure the transformation has not changed the underlying data.
A final challenge is interpreting a possible difference between the overall and city-level results. An airline could appear to perform better overall while showing a different pattern when the destinations are examined separately. I will avoid drawing conclusions from the overall percentages alone and will use the destination-level percentages to explain any discrepancy.