NumPy Statistics and Aggregations

An aggregation turns many numbers into one: the total, the average, the highest. NumPy has a fast function for each, and the axis argument lets you do it per row or per column.

The everyday statistics#

import numpy as np

sales = np.array([120, 95, 180, 150, 95, 210, 170])
print(sales.sum(), sales.mean())
print(np.median(sales))
print(sales.min(), sales.max())
print(round(sales.std(), 1))

Output:

1020 145.71428571428572
150.0
95 210
40.9

• mean: the average. median: the middle value when sorted; it ignores a few extreme values, so it’s often a fairer “typical” number.
• std (standard deviation): how spread out the values are. Small = consistent, large = up and down.

Where is the maximum? argmax#

max() tells you the value; argmax() tells you its position, which is often the more useful answer:

import numpy as np

days = np.array(["Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun"])
sales = np.array([120, 95, 180, 150, 95, 210, 170])
print(sales.argmax(), days[sales.argmax()])
print(days[sales.argmin()])

Output:

5 Sat
Tue

The best day was Saturday. With a tie (95 on Tuesday and Friday), argmin returns the first one.

axis: per column or per row#

For a 2-D table, axis=0 aggregates down each column and axis=1 across each row:

import numpy as np

# rows = students, columns = Maths, Science, English
marks = np.array([[78, 85, 69],
                  [92, 88, 95],
                  [64, 70, 81]])
print(marks.mean(axis=0))
print(marks.mean(axis=1).round(1))

Output:

[78.         81.         81.66666667]
[77.3 91.7 71.7]

• axis=0 → one average per subject (column).
• axis=1 → one average per student (row).

Percentiles#

import numpy as np

scores = np.arange(1, 101)
print(np.percentile(scores, [25, 50, 90]))

Output:

[25.75 50.5  90.1 ]

The 90th percentile means 90% of the values are below it. Exam rank lists and salary reports use this a lot.

Mini project: class results#

Put everything together: who topped, which subject was hardest, and how many students passed every subject.

import numpy as np

names = np.array(["Asha", "Rahul", "Meera", "Kabir", "Zoya"])
subjects = np.array(["Maths", "Science", "English"])
marks = np.array([[78, 85, 69],
                  [92, 88, 95],
                  [64, 70, 81],
                  [45, 58, 62],
                  [88, 79, 91]])

total = marks.sum(axis=1)
print("Topper:", names[total.argmax()], total.max())
print("Hardest subject:", subjects[marks.mean(axis=0).argmin()])
passed_all = (marks >= 50).all(axis=1)
print("Passed every subject:", names[passed_all])

Output:

Topper: Rahul 275
Hardest subject: Maths
Passed every subject: ['Asha' 'Rahul' 'Meera' 'Zoya']

(marks >= 50).all(axis=1) checks, for each student (row), whether all their marks are 50 or more. Kabir scored 45 in Maths, so he’s not in the list.

What’s next#

You now know the core of NumPy. Next, Pandas builds on these arrays and adds column names, missing-value handling and easy loading of CSV and Excel files.

Summary#

Question	Code
Total / average / middle	a.sum(), a.mean(), np.median(a)
Spread	a.std()
Position of best / worst	a.argmax(), a.argmin()
Per column / per row	axis=0 / axis=1
Every / any value matches	mask.all(), mask.any()