agg() 메서드

Calling sum() or mean() directly on a GroupBy object gives you a single statistic. But real analyses often need multiple statistics at once — for example, the total revenue, average order size, and order count per region. The agg() method (short for aggregate) lets you compute all of these in a single, efficient call instead of running separate aggregations and merging results.

The simplest use of agg() is to pass a list of function name strings. Pandas applies every function to the selected column and returns a DataFrame where each column corresponds to one function. This approach works with any built-in aggregation name: 'sum', 'mean', 'min', 'max', 'count', 'std', 'median', and others.

When you pass a list of strings, the output columns are named after the functions themselves ('sum', 'mean' etc.). For cleaner output, use named aggregations: pass keyword arguments where the key is your desired column name and the value is a tuple of (column, function). This is the modern Pandas approach and produces self-documenting code.

You can pass a dictionary to agg() where each key is a column name and each value is a function (or list of functions) to apply to that column. This lets you compute, for example, sum on revenue but mean on units, all in a single GroupBy call.

When you pass a dict with multiple functions per column, the result has a MultiIndex on columns. The first level is the original column name and the second level is the function name. You can flatten these column names into a single string with a list comprehension, making the result easier to work with downstream.

In addition to string names, you can pass any Python callable to agg(). The function receives a Series (the values for that column in one group) and must return a scalar. You can pass a lambda or a named function. Custom functions are more flexible but slower than built-in named functions because they cannot use the C-level optimisations.

Named aggregation syntax also works with callable functions, not just string names. Simply pass a tuple of (column, function) as the keyword argument value, where the function is any callable that accepts a Series and returns a scalar. This keeps your code readable even when mixing built-ins with custom logic.

When you call agg() on a GroupBy without selecting a specific column, Pandas applies the function to every numeric column in the DataFrame. This is a quick way to get a summary of all numeric columns at once. Be aware that columns with non-numeric data types will be silently skipped in most aggregations.

After agg(), the grouping column(s) are the index. A common pattern is to call reset_index() immediately to get a flat DataFrame where the group keys are regular columns. You can then rename, sort, and filter the result like any other DataFrame, making it easy to pass downstream or export.

It is important to distinguish three GroupBy methods: agg() reduces each group to a scalar, producing a result with fewer rows than the original. transform() returns an array with the same shape as the input, so you can add group-level statistics as new columns. apply() is the most flexible but also the slowest, and is covered in the next lesson.

Here is a realistic end-to-end example: compute a sales summary table with total revenue, average order value, number of orders, and the revenue range per region, with clean column names, sorted by total revenue descending. This pattern is directly applicable in product, finance, and marketing analytics workflows.

Test your understanding of the agg() method from this lesson.

In this lesson you learned: how to compute multiple aggregations at once with agg(), how to use named aggregations for clean column names, and how to apply different functions to different columns using a dict. Next up we explore transform() and filter(), which add group-level information back to the original DataFrame.