Data Visualization for Beginners, with Seaborn#

• Seaborn is a Python charting library built on top of matplotlib, specialized for statistical charts like bar plots, box plots, violin plots, and correlation heatmaps.
• This lesson runs entirely inside a Jupyter Notebook in VS Code, one cell at a time.
• No prior seaborn experience is required, only basic Python, pandas, and ideally the matplotlib basics from the previous lesson.
• We will reuse the same real-world restaurant tips dataset from the matplotlib lesson, so you can compare how each library approaches the same questions.

Before You Start#

• Open a new Jupyter Notebook in VS Code and select your Python interpreter as the kernel.
• If seaborn, pandas, or matplotlib are not installed yet, open a terminal in VS Code and run: pip install seaborn pandas matplotlib
• Place tips.csv in the same folder as your notebook so pandas can find it with just its file name.

Why Seaborn, If We Already Have Matplotlib?#

• Matplotlib gives you full control, but statistical charts like box plots or correlation heatmaps take many lines of code to build from scratch.
• Seaborn adds ready-made chart types that understand pandas DataFrames directly, plus attractive default styling, in far fewer lines.
• Under the hood, seaborn is still drawing on matplotlib figures, so everything you learned about plt.xlabel, plt.title, and plt.show still applies.

Bar Plots: Averages by Category#

• seaborn's barplot draws one bar per category, showing a summary statistic, the mean by default, along with a small error bar.

Countplot: How Many Rows per Category#

• countplot is barplot's simpler cousin: instead of averaging a number, it just counts how many rows fall into each category.

A Quick Look at Color Palettes#

• Seaborn ships with several built-in color palettes you can preview before using them in a real chart.

Distribution Plots: histplot and jointplot#

• histplot is seaborn's own histogram, and can overlay a smooth density curve on top with a single argument.
• jointplot goes a step further, combining a scatter plot with a histogram for each axis, all in one figure.

Box Plots: Distribution Summaries by Category#

• A box plot summarizes a number's spread using five landmarks: minimum, 25th percentile, median, 75th percentile, and maximum, plus outlier points.

Violin Plots: Box Plots With Shape#

• A violin plot shows the same summary information as a box plot, but also draws the full smoothed shape of the distribution, so you can see if it is lumpy or has more than one peak.

Strip Plots: Every Individual Point#

• A strip plot plots one dot per row along a category, so instead of a summary shape, you see every single observation.

Swarm Plots: Points That Never Overlap#

• swarmplot works like stripplot, but instead of random jitter, it nudges points apart using an algorithm that guarantees no two points ever overlap.

Combining Plots: Violin Plus Strip#

catplot: One Function, Many Chart Types#

• catplot is a flexible wrapper: change its kind argument, and it draws a completely different chart type using the exact same data and column arguments.

Small Multiples: Faceting With col#

relplot and lineplot: Relationships Between Numbers#

• relplot is scatter and line plots' equivalent of catplot: one flexible, figure-level function covering both.
• lineplot is especially powerful because it automatically summarizes repeated x-values, drawing both an average line and a shaded confidence band.

Pairplot: Every Numeric Column Against Every Other#

• pairplot automatically builds a grid comparing every numeric column in your data against every other numeric column, all in one call.

Regression Plots: Adding a Trend Line#

• regplot draws a scatter plot plus a best-fit straight line through the points, summarizing the overall trend in a single line of code.

Correlation Heatmaps#

• A correlation coefficient measures how strongly two numbers move together, from -1, a perfect opposite relationship, through 0, no relationship, to 1, a perfect matching relationship.
• A heatmap turns a whole table of correlations into a single, easy-to-scan, color-coded grid.

Pivot Tables and Heatmaps Together#

A Finishing Touch: despine#

• Many charts look cleaner with the top and right border lines removed, since they rarely carry any information.

A Real Gotcha: Seaborn's API Changes Over Time#

• Older seaborn tutorials online often use a function called factorplot. It was renamed years ago, and calling it today raises an error.
• The next cell deliberately triggers this exact error on purpose, so you instantly recognize it if you ever run into an old tutorial or old code.

The Fix: Use catplot Instead#

• Whenever you see factorplot in an old tutorial, mentally replace it with catplot, and the rest of the code usually still works unchanged.

Catching Errors Gracefully#

• Sometimes you want to detect a problem and handle it in your code, rather than letting the whole notebook stop.
• Python's try and except lets you catch a specific error type and respond to it instead of crashing.

Best Practice: A Reusable Category Explorer#

• If you find yourself building the same box-plot-by-category chart for many different columns, wrap the pattern in a function.

End-to-End Mini Project: When Do People Tip the Most?#

• Let's combine grouping, a pivot table, and a heatmap to answer a real question: does the average tip percentage change by day and by meal time?

Wrap-Up: What You Learned#

• Bar plots and count plots for comparing averages and counts across categories.
• Color palettes, including accessible, colorblind-friendly options.
• Distribution plots, histplot and jointplot, for seeing shape and relationship together.
• Box plots, violin plots, strip plots, and swarm plots for comparing distributions and individual points, including splitting them by a second category.
• catplot for one flexible function covering many chart kinds at once, including faceting a grid of small multiples with col.
• relplot and lineplot for relationships between numbers, including automatic confidence bands.
• pairplot for a fast overview of every numeric column against every other.
• Regression plots, regplot and lmplot, for adding trend lines and faceting by category.
• Correlation heatmaps and pivot-table heatmaps for spotting patterns in a full table of numbers at a glance.
• Small finishing touches, like despine, that make any chart look more polished.
• Practice prompt: pick a dataset of your own and try rebuilding this lesson's end-to-end mini project a pivot table turned into a heatmap to answer your own question.
• If this lesson helped, consider subscribing for more hands-on data tutorials and drop a comment with which chart type you want to see covered next!