Checkpoint: Analysis & Aggregation

Reviewed & published by Brayan K

This checkpoint is a hands-on review that combines every analysis skill from this section — binning, encoding, windows, group transforms, time-series math, and chaining — into one build challenge and a short self-test quiz.

Part of the free Pandas course at LearnCodingFast — hands-on lessons with examples you run in your browser, plus practice exercises and a quick quiz.

Work through the multi-step challenge first, check your answer against the full solution, then test your recall with the quiz before moving on to the capstone.

What You've Learned in This Section

• Binning with pd.cut and pd.qcut
• One-hot encoding with get_dummies
• Rolling, expanding & ewm windows
• GroupBy transform and filter
• Time-series shift, diff & pct_change
• Plotting, method chaining & styling

1 Quick Recap — The Right Tool For The Job

Before the challenge, here is the mental map of when to reach for each tool you have met:

• pd.cut / pd.qcut — turn a continuous column into labelled buckets (fixed edges vs equal frequency).
• pd.get_dummies — expand a category column into 0/1 indicator columns for modelling.
• .rolling / .expanding / .ewm — moving averages, running totals, recency-weighted means.
• groupby().transform — broadcast a group statistic back onto every row (same shape).
• groupby().filter — keep or drop whole groups by a predicate.
• shift / diff / pct_change — compare each row to an earlier one for trends and returns.

2 Warm-Up — Bin, Roll, Transform

A quick three-skill warm-up: bin a column with cut, add a rolling average, and broadcast a group mean with transform. Run it and read each new column.

# bin -> rolling -> transform, all on one frame
df["band"]   = pd.cut(df["sales"], bins=[0, 50, 100], labels=["lo", "hi"])
df["ma2"]    = df["sales"].rolling(2).mean()
df["b_mean"] = df.groupby("band", observed=True)["sales"].transform("mean")

import pandas as pd

df = pd.DataFrame({"day": [1, 2, 3, 4],
                   "sales": [30, 80, 40, 90]})

df["band"]   = pd.cut(df["sales"], bins=[0, 50, 100], labels=["lo", "hi"])
df["ma2"]    = df["sales"].rolling(2).mean()
df["b_mean"] = df.groupby("band", observed=True)["sales"].transform("mean")
print(df)

# ✅ Expected output:
#    day  sales band   ma2  b_mean
# 0    1     30   lo   NaN    35.0
# 1    2     80   hi  55.0    85.0
# 2    3     40   lo  60.0    35.0
# 3    4     90   hi  65.0    85.0

3 The Build Challenge — A Sales Analysis Pipeline

You are given a daily sales table per region. Build a single analysis that does all of the following, in order:

• Bin sales into "low" / "high" at the edge 60 with pd.cut
• Add a 2-day rolling average column roll2
• Add a within-region z-score column z using transform
• Summarise: print total sales per region

Try it in the editor first. The full reference solution is hidden just below — open it only after you have attempted the build.

import pandas as pd

df = pd.DataFrame({
    "day":    [1, 2, 3, 4, 5, 6],
    "region": ["N", "N", "N", "S", "S", "S"],
    "sales":  [40, 70, 55, 90, 30, 80],
})

# 1) bin sales at edge 60 -> "low" / "high"
# df["band"] = pd.cut(...)

# 2) 2-day rolling average -> df["roll2"]

# 3) within-region z-score -> df["z"]
#    def z(x): return (x - x.mean()) / x.std()

# 4) print total sales per region
# print(df.groupby("region")["sales"].sum())

print(df)

# ✅ Expected output:
#    day region  sales
# 0    1      N     40
# 1    2      N     70
# 2    3      N     55
# 3    4      S     90
# 4    5      S     30
# 5    6      S     80

import pandas as pd

df = pd.DataFrame({
    "day":    [1, 2, 3, 4, 5, 6],
    "region": ["N", "N", "N", "S", "S", "S"],
    "sales":  [40, 70, 55, 90, 30, 80],
})

# 1) bin sales into low / high
df["band"] = pd.cut(df["sales"],
                    bins=[0, 60, 200],
                    labels=["low", "high"])

# 2) 2-day rolling average
df["roll2"] = df["sales"].rolling(2).mean()

# 3) within-region z-score with transform
def z(x):
    return (x - x.mean()) / x.std()

df["z"] = df.groupby("region")["sales"].transform(z).round(3)

print(df)
#    day region  sales  band  roll2      z
# 0    1      N     40   low    NaN -1.000
# 1    2      N     70  high   55.0  1.000
# 2    3      N     55   low   62.5  0.000
# 3    4      S     90  high   72.5  0.726
# 4    5      S     30   low   60.0 -1.141
# 5    6      S     80  high   55.0  0.415

# 4) total sales per region
print(df.groupby("region")["sales"].sum())
# region
# N    165
# S    200
# Name: sales, dtype: int64

! Common Errors (And How to Fix Them)

❌ Assigning an agg result as a column

agg shrinks the frame and misaligns on assignment:

df["z"] = df.groupby("region")["sales"].mean()   # ❌ wrong shape

✅ Fix: use transform to keep the row count:

df["z"] = df.groupby("region")["sales"].transform("mean")   # ✅

❌ Rolling across groups by accident

df["roll2"] = df["sales"].rolling(2).mean()   # ❌ blends regions at the boundary

✅ Fix: roll inside each group when needed:

df.groupby("region")["sales"].rolling(2).mean()   # ✅

📝 Checkpoint Quiz

Answer each in your head, then expand to check.

pd.qcut. It splits on quantiles so every bin is equally full, unlike pd.cut which uses fixed value edges.

dtype=int. And drop_first=True removes the redundant first indicator column.

A full window of 3 is not available yet for the first two rows. Pass min_periods=1 to compute with partial windows instead.

transform returns the same shape as the input (one value per row, broadcast within each group), while agg collapses each group to a single summary row.

diff() is the absolute change in units; pct_change() is the relative change as a fraction (a 0.25 means +25%).

It returns all rows of the groups that pass the test. The function receives each group as a sub-DataFrame and must return a single boolean (True keeps the whole group).

❓ Frequently Asked Questions

Checkpoint cleared — you can analyse like a pro!

You combined binning, rolling windows, and group transforms into one pipeline, summarised by group, and tested your recall. That is the full analysis toolkit working together.

Practice quiz

Continue this course

• Previous: Styling & Formatting Output
• Next: Time Series Resampling (resample, asfreq) — Downsample and upsample time series with resample and asfreq
• Quick reference: Pandas cheat sheet

Frequently asked questions

What should I be able to do after this checkpoint?

You should be comfortable binning a numeric column, one-hot encoding categories, computing rolling and expanding windows, broadcasting group statistics with transform, filtering whole groups, deriving time-series changes with shift/diff/pct_change, and stitching it together with method chaining — then presenting the result cleanly.

Do I need to memorise every method signature?

No. The goal is to recognise which tool fits which problem: cut/qcut for binning, get_dummies for encoding, rolling/expanding for windows, transform/filter for groups, shift/diff/pct_change for time series. Looking up exact arguments is normal and expected.

Why combine transform and binning in one challenge?

Real analysis rarely uses a single method. Binning turns a continuous column into segments, and transform adds a within-segment statistic back onto every row, so the two together let you compare each record against its own bucket — a very common analytics pattern.

What comes after this checkpoint?

The capstone: an end-to-end data cleaning and analysis project that pulls together everything from the course — reading data, cleaning it, wrangling it, aggregating it, and presenting the result. This checkpoint is your warm-up for that.