Introduction

This will look at delays in domestic flight data in January 2016. As with all data, the data first had to be read and checked for missing data. To ensure that missing data was due only for cancelled or diverted flights, the count for data not explained by either of those reasons was set to zero. Additionally,the dates were changed from a character class into a date class and the data was formatted so that times would be uniform with 4 digits and the POSIX class was used to deal with timezone manipulation. The cancelled or diverted flights were filtered out as they were not relevant.
Flights <- read.csv("/Users/wendyhayes/Desktop/MBA 676-R/domestic_flights_jan_2016.csv", stringsAsFactors = FALSE)
library(dplyr)
library(knitr)
library(ggvis)
IncompleteCases <- Flights %>% filter(!complete.cases(.)) 
length(IncompleteCases[IncompleteCases$Cancelled==0 & IncompleteCases$Diverted==0])
## [1] 0
Flights$FlightDate <- as.Date(Flights$FlightDate, format = "%m/%d/%Y")

Flights <- Flights %>% mutate(New_CRSDepTime = paste(FlightDate, sprintf("%04d", CRSDepTime)), New_CRSArrTime = paste(FlightDate, sprintf("%04d", CRSArrTime)), New_DepTime = paste(FlightDate, sprintf("%04d", DepTime)), New_ArrTime = paste(FlightDate, sprintf("%04d", ArrTime)), New_WheelsOff = paste(FlightDate, sprintf("%04d", WheelsOff)), New_WheelsOn = paste(FlightDate, sprintf("%04d", WheelsOn)))

Flights$New_CRSDepTime <- as.POSIXct(Flights$New_CRSDepTime, format="%Y-%m-%d %H%M")
Flights$New_CRSArrTime <- as.POSIXct(Flights$New_CRSArrTime, format="%Y-%m-%d %H%M")
Flights$New_DepTime <- as.POSIXct(Flights$New_DepTime, format="%Y-%m-%d %H%M")
Flights$New_ArrTime <- as.POSIXct(Flights$New_ArrTime, format="%Y-%m-%d %H%M")
Flights$New_WheelsOff <- as.POSIXct(Flights$New_WheelsOff, format="%Y-%m-%d %H%M")
Flights$New_WheelsOn <- as.POSIXct(Flights$New_WheelsOn, format="%Y-%m-%d %H%M")

Flights <- Flights %>% filter(Cancelled == 0, Diverted == 0) %>% filter(ArrTime>DepTime)

Delays: Length of Time and Location

The difference between the actual and scheduled arrival times were calculated and only those with a delay were examined as that information is most important to a traveler. A histogram showed that most of the delayed flights still seem to arrive within an hour or two of the scheduled arrival time. Since the data covers January and includes locations such as Chicago and Denver, delays may be expected.
Flights <- Flights %>% filter(Cancelled == 0) %>%
     mutate(ArrDelay = as.integer(New_ArrTime - New_CRSArrTime))
Flights<- Flights [Flights$ArrDelay>=0,] 
Flights$ArrDelay <- Flights$ArrDelay/60 

Flights %>% ggvis(~Flights$ArrDelay) %>% layer_histograms(width=50, fill := "green") %>%
  add_axis("y", title = "Frequency of Delays", title_offset = 80) %>%
  add_axis("x", title = "Minutes Delayed")
percent <- seq(0,.9,.10)
quants <- 1-percent
delayvalues <- quantile(Flights$ArrDelay, quants)
dvalues <- as.numeric(delayvalues)
qdif <- data.frame (MinutesDelayed = dvalues, Probability = quants)
pcity <- ggvis(qdif,x=~MinutesDelayed, y= ~Probability)
layer_points(pcity, fill = "blue")
Looking at quantiles of the delays depicted the relationship between the length of the delay and the probability of experiencing that delay. A point plot seemed to best that relationship. In other words, would an airplane be likely to be only a few minutes late or was it likely to be many hours late.The plot demonstrated that 90 percent of the flights are under 2 hours delayed. The next thing determined was which locations had the most delays. The charts below display the top twenty cities of origin and the top twenty destination cities with the greatest median arrival delays in minutes. The last chart shows the cities that, both as the origin and the destination, have large arrival delays. The bar plot shows the cities involved with arrival delays as both the origin and destination.
medbycity <- by(Flights$ArrDelay, Flights$DestCityName,median)
TopTwentyDelayed <- sort(medbycity, decreasing=TRUE)[1:20]
DelayedDepart <- data.frame(DestinationCity=rownames(TopTwentyDelayed),DestinationMedianDelayed=as.numeric(TopTwentyDelayed))
kable(DelayedDepart)
DestinationCity DestinationMedianDelayed
Plattsburgh, NY 131.0
Fairbanks, AK 61.0
North Bend/Coos Bay, OR 47.0
Rock Springs, WY 40.0
Cody, WY 36.5
Alpena, MI 35.0
Santa Maria, CA 33.5
Aguadilla, PR 32.0
Bend/Redmond, OR 28.0
Trenton, NJ 27.5
Muskegon, MI 27.0
Roswell, NM 27.0
San Francisco, CA 27.0
Bellingham, WA 26.0
Latrobe, PA 25.5
Christiansted, VI 24.0
Paducah, KY 24.0
Adak Island, AK 23.5
Devils Lake, ND 23.0
San Luis Obispo, CA 23.0 
MedOriginCity <- by(Flights$ArrDelay,Flights$OriginCityName,median)
DelayedOrigin <- sort(MedOriginCity, decreasing =TRUE)[1:20]
DelayedbyOrigin <- data.frame(OriginCity=rownames(DelayedOrigin),OriginMedianDelayed=as.numeric(DelayedOrigin))
kable(DelayedbyOrigin)
OriginCity OriginMedianDelayed
Escanaba, MI 95.5
Cordova, AK 83.0
Yakutat, AK 67.0
Santa Maria, CA 38.0
Iron Mountain/Kingsfd, MI 32.0
Key West, FL 31.0
Meridian, MS 31.0
Niagara Falls, NY 31.0
North Bend/Coos Bay, OR 31.0
Columbus, GA 30.0
Eau Claire, WI 28.5
Eugene, OR 28.0
Daytona Beach, FL 27.0
Medford, OR 27.0
Muskegon, MI 27.0
Sitka, AK 27.0
Brownsville, TX 26.0
Plattsburgh, NY 26.0
Montgomery, AL 25.0
Arcata/Eureka, CA 24.0 
DelayOrigandDest <- inner_join(TopDelayed,DelayedbyOrigin,by=c("DestinationCity" = "OriginCity"))
 colnames(DelayOrigandDest) <- c("City","MedianDestDelay","MedianOriginDelay") 


kable(DelayOrigandDest)
City MedianDestDelay MedianOriginDelay
Plattsburgh, NY 131.0 26
North Bend/Coos Bay, OR 47.0 31
Santa Maria, CA 33.5 38
Muskegon, MI 27.0 27 
PlotDataTogether <- data.frame(City=rep.int(DelayOrigandDest$City,2),Delay=c(DelayOrigandDest$MedianDestDelay, DelayOrigandDest$MedianOriginDelay), DelayType=c(rep.int("Destination",4),rep.int("Origination",4)))
Bothplotted <- ggvis(PlotDataTogether,x=~City, y= ~Delay, fill = ~factor(DelayType))
layer_bars(Bothplotted)

Conclusion

It looks like the smaller airports had the longest delay times. Also, if you’re planning on flying to or from Plattsburgh, NY, bring a good book to read because there’s a good chance there will be extra time.