The describe() function generates statistics for numeric columns. This function can be used in two modes:
- Regular Aggregate Mode
It computes the count, mean, std, min, percentiles, and max for numeric columns.
Default statistics include: "count", "mean", "std", "min", "percentile", "max".
If describe() is used on the output of any DataFrame API or groupby(), then it is used in regular aggregate mode. - Time Series Aggregate Mode
It computes max, mean, min, std, median, mode, and percentiles for numeric columns.
Default statistics include: 'max', 'mean', 'min', 'std'
If describe() is used on the output of groupby_time(), then it is used in time series aggregate mode, where time series aggregates are used to calculate the statistics.
Examples here are for describe() as regular function or aggregate function. For describe() as Time Series Aggregate, refer to describe() in Time Series Aggregate Mode.
Example Prerequisite
>>> df = DataFrame('sales')
>>> df
Feb Jan Mar Apr datetime
accounts
Alpha Co 210.0 200 215 250 04/01/2017
Red Inc 200.0 150 140 None 04/01/2017
Orange Inc 210.0 None None 250 04/01/2017
Jones LLC 200.0 150 140 180 04/01/2017
Yellow Inc 90.0 None None None 04/01/2017
Blue Inc 90.0 50 95 101 04/01/2017
Example: Generates statistics for DataFrame "sales"
Use default values to computes count, mean, std, min, percentiles, and max for numeric columns.
>>> df.describe()
Apr Feb Mar Jan
func
count 4 6 4 4
mean 195.25 166.667 147.5 137.5
std 70.971 59.554 49.749 62.915
min 101 90 95 50
25% 160.25 117.5 128.75 125
50% 215 200 140 150
75% 250 207.5 158.75 162.5
max 250 210 215 200
Example: Use argument percentiles to compute the 30th and 60th percentiles
>>> df.describe(percentiles=[.3, .6])
Apr Feb Mar Jan
func
count 4 6 4 4
mean 195.25 166.667 147.5 137.5
std 70.971 59.554 49.749 62.915
min 101 90 95 50
30% 172.1 145 135.5 140
60% 236 200 140 150
max 250 210 215 200
Example: Use groupby to compute statistics for specific groups
>>> df1 = df.groupby(["datetime", "Feb"])
>>> df1.describe()
Jan Mar Apr
datetime Feb func
04/01/2017 90.0 25% 50 95 101
50% 50 95 101
75% 50 95 101
count 1 1 1
max 50 95 101
mean 50 95 101
min 50 95 101
std None None None
200.0 25% 150 140 180
50% 150 140 180
75% 150 140 180
count 2 2 1
max 150 140 180
mean 150 140 180
min 150 140 180
std 0 0 None
210.0 25% 200 215 250
50% 200 215 250
75% 200 215 250
count 1 1 2
max 200 215 250
mean 200 215 250
min 200 215 250
std None None 0
Example: Use argument include value 'all' to compute statistics for all columns
Computes count, mean, std, min, percentiles, and max for numeric columns and computes count and unique for non-numeric columns.
>>> df.describe(include="all")
accounts Feb Jan Mar Apr datetime
func
25% None 117.5 125 128.75 160.25 None
75% None 207.5 162.5 158.75 250 None
count 6 6 4 4 4 6
mean None 166.667 137.5 147.5 195.25 None
max None 210 200 215 250 None
min None 90 50 95 101 None
50% None 200 150 140 215 None
std None 59.554 62.915 49.749 70.971 None
unique 6 None None None None 1