Overview

This data set includes information from two airlines: ALASKA and AM WEST. These two airlines have 6 destinations that they operate with and this data includes a count of flights that were on time and delayed for each of the 6 cities.

data <- read.csv("https://raw.githubusercontent.com/addsding/data607/main/assignment5/data.csv")
head(data)

Tidying the Data

Tidy data is data where:

We see that to start with, every row here contains two observations: On Time and Delayed. We first want to separate this into two rows in order to have clean data.

data_pivoted <- pivot_longer(data, cols=3:4, names_to="arrival_type", values_to="flight_count")
head(data_pivoted)

Looks good! Next, it looks like there’s a . in On.Time – I’ll replace those with a space instead.

data_pivoted$arrival_type <- gsub("\\.", " ", data_pivoted$arrival_type)
head(data_pivoted)

The data is now clean for analysis!

Analysis

First, let’s see just nominally how many flights were delayed for each airline.

total_delayed <- data_pivoted |>
  group_by(Airline) |>
  summarise(total_delayed = sum(flight_count[arrival_type=="Delayed"]), .groups = 'drop') |>
  arrange(total_delayed)

head(total_delayed)

We can see here that nominally, AM WEST has more delays. What percent of flights is that though?

delayed_percentage <- data_pivoted |>
  group_by(Airline) |>
  summarise(total_delayed = sum(flight_count[arrival_type=="Delayed"]),
            total_flights = sum(flight_count),
            delayed_percent = sum(flight_count[arrival_type=="Delayed"])
              /sum(flight_count),
            .groups = 'drop') |>
  arrange(total_delayed, total_flights, delayed_percent)

head(delayed_percentage)

Despite having more delays nominally, ALASKA actually has a higher delay rate (13.3% vs. 10.9%). Do we see any city that perhaps is driving up delays for both or either of the airlines?

by_city <- data_pivoted |>
  group_by(Airline, Destination) |>
  summarise(total_delayed = sum(flight_count[arrival_type=="Delayed"]),
            total_flights = sum(flight_count),
            delayed_percent = sum(flight_count[arrival_type=="Delayed"])
              /sum(flight_count),
            .groups = 'drop') |>
  arrange(desc(delayed_percent), total_delayed, total_flights)

# round delayed_percent 
by_city <- by_city |> mutate(across(delayed_percent, round, 2))

# bar chart for delay rates
delay_bar <-ggplot(data=by_city
                   , aes(x=Destination
                         , y=delayed_percent
                         , fill=Airline)) + 
  geom_bar(stat="identity", position=position_dodge()) + 
  geom_text(aes(label=delayed_percent), vjust=1.6, color="white",
            position = position_dodge(0.9), size=3.5)+
  scale_fill_brewer(palette="Paired")+
  theme_minimal()

# bar chart for counts of flights
flight_bar <-ggplot(data=by_city
                   , aes(x=Destination
                         , y=total_flights
                         , fill=Airline)) + 
  geom_bar(stat="identity", position=position_dodge()) + 
  geom_text(aes(label=total_flights), vjust=1.6, color="white",
            position = position_dodge(0.9), size=3.5)+
  scale_fill_brewer(palette="Paired")+
  theme_minimal()


delay_bar

flight_bar

We can see here that for AM WEST, San Francisco and Seattle are huge outliers with delay rates higher than 20%. The average for AM WEST is 10% though – it seems like Phoenix really brings the rate down as only 8% of flights to that destination are delayed and a vast majority of flights go there (5k).

For ALASKA, San Francisco is the most extreme at 17% and the only other one above the average delay rate (13.3%) is Seattle at 14%. ALASKA’s flights mainly seem to go to Seattle (2k) so the overall average being close to Seattle’s average makes sense.

Overall, AM WEST seems to have a wider range of delay rates across the different destinations (min of 8, max of 29) while ALASKA’s numbers are less diverse (min of 5, max of 17). The variability of AM WEST seems to rely heavily on what cities the flight is going to and it may be due to less experience flying to that region as a majority of flights go towards Phoenix and that’s its least delayed destination. ALASKA on the other hand still has Seattle (its most popular destination) as its second most delayed flight which signals that there probably be some sort of improvement. Both airlines seem to struggle the most with Seattle and San Francisco.

In the end, AM WEST is more delayed for every destination – if I were to pick one of these airlines, I would likely pick AM WEST for less of a chance of being delayed.

Conclusion

The tidying and analysis of this data using tidyr and dplyr was overall made very simple with how extensive and easy-to-use each library is. For future analysis ideas, having this data broken down by any sort of time period (weeks, months, years, etc.) or adding weather conditions could help narrow down as to why certain flights were delayed or not.