GroupBy

What is groupby?

• A method to split data into groups based on some key(s).
• Perform aggregations, transformations, or custom functions on each group.
• Combine results into a new DataFrame or Series.

Basic Syntax

df.groupby(by)[column].operation()

• by → column(s) or index level(s) to group on.
• operation → aggregation (sum, mean, count, etc.), transformation, or apply.

Aggregation

Common built-ins:

• .mean() → average per group
• .sum() → sum per group
• .count() → count non-NA entries per group
• .size() → group sizes (including NAs)
• .max() / .min() → extrema per group
• .median(), .std(), .var()

Example:

import pandas as pd​data = {    "Department": ["Sales", "Sales", "IT", "IT", "HR"],    "Salary": [50000, 60000, 70000, 80000, 45000]}df = pd.DataFrame(data)​df.groupby("Department")["Salary"].mean()

Output:

DepartmentHR        45000IT        75000Sales     55000Name: Salary, dtype: int64

Multiple Aggregations

Use .agg() for flexibility:

df.groupby("Department")["Salary"].agg(["mean", "sum", "max"])

Or custom names:

df.groupby("Department")["Salary"].agg(    avg_salary="mean", total_salary="sum")

Grouping by Multiple Columns

df.groupby(["Department", "Salary"]).size()

This produces a MultiIndex.

Transformations

Unlike aggregation (reduces groups), transformations return the same shape as the original DataFrame.
Example: normalise salaries within departments:

df["Salary_normalized"] = df.groupby("Department")["Salary"].transform(lambda x: x / x.mean())

Filtering Groups

Use .filter() to drop groups based on a condition:

df.groupby("Department").filter(lambda x: x["Salary"].mean() > 60000)

Keeps only departments with an average salary above 60,000.

Iterating Over Groups

for name, group in df.groupby("Department"):    print(name)    print(group)

Returns multiple DataFrames:

• name → the group key (value of the column you grouped by).
• group → a DataFrame containing only the rows that belong to that group.

If you group by multiple columns, name becomes a tuple of keys.