Departure delays

2 500 scheduled domestic departures in June–August 2024 operated by four fictional airlines from five fictional airports. Each row is one flight: carrier, origin, distance, scheduled departure hour, whether adverse weather was reported, and the departure delay in minutes (negative = early). Delays are mostly small with a long right tail, so the data are made for medians, nonparametric tests (Mann–Whitney, Kruskal–Wallis) and robust summaries.

Open in the lab Download CSV

Rows: 2,500
Columns: 8
One row is: one scheduled flight departure
Version: v1

Data dictionary

Column	Type	Unit	Summary	Description
`flight_id` Flight	id			Flight record identifier.
`date` Date	date			Scheduled departure date.
`carrier` Airline	category		4 levels: Amber Air, Cobalt Wings, Jade Skyways, Sienna Jet	Operating airline (fictional).
`origin` Origin	category		5 levels: North Field, Harbour, Central, Lakeside, Ridge	Departure airport (fictional).
`distance_km` Distance	integer	km	mean 637.2, median 586, SD 310.3, range 180–1500	Great-circle flight distance.
`dep_hour` Scheduled hour	integer	hour	mean 14.48, median 15, SD 5.281, range 6–23	Scheduled departure hour (0–23).
`weather` Adverse weather	boolean		2 levels	True if adverse weather (dust storm, thunderstorm, low visibility) was reported at departure.
`delay_min` Departure delay	integer	minutes	mean 21.16, median −1, SD 48.76, range −17–509	Actual minus scheduled departure time; negative = early.

What it is good for

Describe a distribution with a spike near zero and a long right tail using the median, IQR and percentiles.
Compare delays across carriers with the Kruskal–Wallis test and follow-up pairwise tests.
Compare weather vs no-weather flights with the Mann–Whitney test.
Explain why the mean delay is pulled far above the typical delay.

Methods: median and IQR, box plot, Mann–Whitney U test, Kruskal–Wallis test, Dunn's test, proportion delayed > 15 minutes.

Data quality notes

delay_min has many values at or below zero and a very long right tail (some flights delayed several hours).
Carriers fly different mixes of hours and airports, so raw carrier comparisons are partly confounded.
No missing values (cancelled flights are not included).

Provenance and licence

Each flight is on time or early (delay = −Gamma, a few minutes early) with a probability that depends on carrier, hour and weather; otherwise its delay is drawn from a heavy-tailed gamma (shape 0.8) whose scale grows with the carrier's delay propensity (Amber 1.0, Cobalt 1.35, Jade 0.8, Sienna 1.15), later departure hours (knock-on delays) and bad weather (×1.8). Delays are rounded to whole minutes and capped at 600.

Synthetic data · CC0-1.0 (synthetic data generated by Statistics Quest) · generated deterministically (seed 20241020). It does not describe real people or organisations; any resemblance is coincidental.

Preview

Dataset

Departure delays

2,500 rows · one row = one scheduled flight departure · synthetic data

About this dataset Open in the lab

Flight	Date	Airline	Origin	Distance (km)	Scheduled hour (hour)	Adverse weather	Departure delay (minutes)
F0001	2024-06-01	Sienna Jet	Central	561	8	false	0
F0002	2024-06-01	Cobalt Wings	Ridge	394	9	false	3
F0003	2024-06-01	Amber Air	Lakeside	451	9	false	7
F0004	2024-06-01	Sienna Jet	North Field	776	10	false	-1
F0005	2024-06-01	Sienna Jet	Ridge	1285	11	false	-3
F0006	2024-06-01	Amber Air	Central	378	12	false	5
F0007	2024-06-01	Cobalt Wings	Central	1155	12	false	12
F0008	2024-06-01	Amber Air	Central	604	13	false	3
F0009	2024-06-01	Amber Air	North Field	353	18	false	-1
F0010	2024-06-01	Cobalt Wings	Lakeside	229	18	true	19
F0011	2024-06-01	Amber Air	Central	870	18	false	-2
F0012	2024-06-01	Sienna Jet	North Field	971	18	false	-11

Used in

Five-number summary and outliers · Describing Data
Box plots and dot plots · Visualising and Exploring Data
Mann–Whitney U test · Comparing Groups and Categorical Data
Kruskal–Wallis test · Comparing Groups and Categorical Data
Choosing a test · Comparing Groups and Categorical Data
Which carrier is more punctual? · Case