This project will utilize the nycflights13 package added to R by Hadley Wickham. It contains information about all flights that departed from NYC in 2013.
library(nycflights13)
library(dplyr)
library(psych)
nycflights <- flights
nycweather <- weather
colnames(nycflights)## [1] "year" "month" "day" "dep_time"
## [5] "sched_dep_time" "dep_delay" "arr_time" "sched_arr_time"
## [9] "arr_delay" "carrier" "flight" "tailnum"
## [13] "origin" "dest" "air_time" "distance"
## [17] "hour" "minute" "time_hour"# We only care about the performances of airports, not individual airlines or airplanes
nycflights <- nycflights[, c(1:9, 13:17)]
# We do not care about dew point, nor humidity, nor the direction of the wind.
nycweather <- nycweather[, c(1:6, 10:15)]
# Combine the datasets into one large dataset
# Then filter any incomplete rows from the merging
nycfinal <- left_join(nycflights, nycweather, by = c("origin", "year", "month", "day", "hour"))
nycfinal <- nycfinal[complete.cases(nycfinal), ]Does weather affect airport delays? And which causes the most delays? Wind, precipitation, or visibility?
Each case is a flight from a NYC airport (Newark, JFK, and LaGuardia) with the weather report for that flight’s departure time. There are 291,675 cases.
According to Hadley Wickham, data is collected from the Bureau of Transportation Statistics.
This is an observational study.
The Bureau of Transportation Statistics for the year 2013.
The response variable is delay and is numerical.
The explanatory variables are wind, precipitation, and visibility. All are numeric.
describe(nycfinal$dep_delay)## vars n mean sd median trimmed mad min max range skew
## X1 1 291675 10.99 37.59 -2 2.38 5.93 -43 1301 1344 5.09
## kurtosis se
## X1 50.69 0.07describe(nycfinal$arr_delay)## vars n mean sd median trimmed mad min max range skew
## X1 1 291675 4.97 42.14 -6 -2.3 20.76 -86 1272 1358 3.84
## kurtosis se
## X1 32.44 0.08describe(nycfinal$wind_gust)## vars n mean sd median trimmed mad min max range skew
## X1 1 291675 12.06 14.25 11.92 11.61 5.89 0 1206.43 1206.43 66.63
## kurtosis se
## X1 5577.33 0.03describe(nycfinal$wind_speed)## vars n mean sd median trimmed mad min max range skew
## X1 1 291675 10.48 12.38 10.36 10.09 5.12 0 1048.36 1048.36 66.63
## kurtosis se
## X1 5577.33 0.02describe(nycfinal$precip)## vars n mean sd median trimmed mad min max range skew kurtosis
## X1 1 291675 0 0.01 0 0 0 0 0.38 0.38 15 283.17
## se
## X1 0describe(nycfinal$visib)## vars n mean sd median trimmed mad min max range skew kurtosis
## X1 1 291675 9.55 1.56 10 9.98 0 0 10 10 -4.11 17.5
## se
## X1 0# On time/early vs delayed departures
# Positive value means delayed flight
table(nycfinal$dep_delay >= 0)##
## FALSE TRUE
## 167455 124220# On time/early vs delayed arrivals
# Positive value means delayed flight
table(nycfinal$arr_delay >= 0)##
## FALSE TRUE
## 173091 118584# Delayed departures
describe(subset(nycfinal, dep_delay > 0)$dep_delay)## vars n mean sd median trimmed mad min max range skew
## X1 1 109330 36.9 51.78 18 25.94 22.24 1 1301 1300 3.62
## kurtosis se
## X1 27.97 0.16# On time/early departures
describe(subset(nycfinal, dep_delay <= 0)$dep_delay)## vars n mean sd median trimmed mad min max range skew kurtosis
## X1 1 182345 -4.54 3.04 -4 -4.35 2.97 -43 0 43 -0.67 0.76
## se
## X1 0.01# Delayed arrivals
describe(subset(nycfinal, arr_delay > 0)$arr_delay)## vars n mean sd median trimmed mad min max range skew
## X1 1 113701 37.92 50.91 19 27.31 20.76 1 1272 1271 3.59
## kurtosis se
## X1 27.56 0.15# On time/early arrivals
describe(subset(nycfinal, arr_delay <= 0)$arr_delay)## vars n mean sd median trimmed mad min max range skew
## X1 1 177974 -16.08 10.89 -15 -15.16 11.86 -86 0 86 -0.8
## kurtosis se
## X1 0.61 0.03The Beaufort Scale categorizes wind types for both land and sea. According to this scale, anything about 32 MPH is qualified as High Wind/moderate gale/near gale.
# High wind according to Beaufort Scale
describe(subset(nycfinal, wind_speed >= 32)$wind_speed)## vars n mean sd median trimmed mad min max range skew
## X1 1 421 113.78 272.89 34.52 34.42 1.71 32.22 1048.36 1016.14 3.13
## kurtosis se
## X1 7.79 13.3# Light wind
describe(subset(nycfinal, wind_gust >= 32)$wind_gust)## vars n mean sd median trimmed mad min max range skew
## X1 1 1360 63.93 180.26 34.43 35.2 1.96 33.11 1206.43 1173.33 6.17
## kurtosis se
## X1 36.16 4.89According to this Ask the Captain, pilots need 600ft or approximately 0.11 miles of visibility for takeoffs. Other sources from Googling report .5 miles.
# Good visibility
describe(subset(nycfinal, visib >= .5)$visib)## vars n mean sd median trimmed mad min max range skew kurtosis
## X1 1 290212 9.59 1.41 10 9.98 0 0.5 10 9.5 -4.18 18.32
## se
## X1 0# Bad visibility
describe(subset(nycfinal, visib < .5)$visib)## vars n mean sd median trimmed mad min max range skew kurtosis se
## X1 1 1463 0.2 0.07 0.25 0.22 0 0 0.25 0.25 -1.26 0.27 0