Categorical Plots: boxplot, barplot, violinplot

When you want to compare a numeric variable across different categories — for example, tips received by day of the week, or product prices by brand — you need specialised plots. Seaborn provides three essential categorical comparison plots: boxplot, barplot, and violinplot. Each one reveals different aspects of the distribution and is suited to different analytical goals.

A box plot summarises a distribution using five statistics: the minimum, first quartile (Q1), median, third quartile (Q3), and maximum (excluding outliers). The box spans Q1 to Q3 (the Interquartile Range, IQR), the line inside the box is the median, and the whiskers extend to 1.5×IQR. Points beyond the whiskers are plotted individually as outliers. Box plots are excellent for spotting skewness and outliers at a glance.

The hue parameter splits each category into sub-groups shown side by side with different colours. For example, adding hue='smoker' to a day-by-bill box plot shows two boxes per day — one for smokers and one for non-smokers. This makes three-variable comparisons clear without needing facets. Use a legend to keep the plot readable.

sns.barplot() shows the mean of a numeric variable for each category, with error bars representing a 95% confidence interval by default (computed via bootstrapping in older Seaborn versions, or from the standard error in newer ones). Unlike histograms, bar plots summarise each group in a single value — they are best when you care about the average and its uncertainty, not the full distribution shape.

A violin plot combines a box plot with a KDE curve mirrored on both sides. The width of the violin at each y-value represents the density of data at that point — wide sections have many observations, narrow sections have few. Violin plots convey more information than box plots because they show whether the distribution is unimodal or multimodal, but they require more data to be meaningful (aim for n > 20 per group).

The inner parameter of violinplot controls what is drawn inside the violin shape. Options include 'box' (a mini box plot), 'quartile' (horizontal lines at Q1, median, Q3), 'point' (individual data points), and 'stick' (vertical tick marks per observation). Setting inner=None draws only the density envelope with no internal marks, giving the cleanest look for side-by-side comparisons.

Adding a sns.stripplot() on top of a box or violin plot shows every individual data point as a dot, preventing hiding of outliers and small sample sizes. Set jitter=True to spread points horizontally so they do not all stack on top of each other, and set alpha to make them semi-transparent when there are many points.

Use a box plot when you need to see medians, IQR, and outliers and sample sizes are moderate. Use a bar plot when your audience cares about means and confidence intervals (common in reports and dashboards). Use a violin plot when you want to reveal the full distributional shape — especially multimodality. For small datasets (n < 30 per group), prefer box plots plus strip plots over violin plots because KDE curves become unreliable with few points.

sns.catplot() is the figure-level wrapper for all categorical plots. Pass kind='box', 'bar', 'violin', 'strip', 'swarm', 'count', or 'point' to switch plot types. Its key advantage is the col and row parameters that create a grid of subplots automatically — for example, separate panels per gender or time of day without manually creating plt.subplots.

sns.countplot() shows how many observations fall into each category — essentially a bar chart of counts. It is equivalent to calling value_counts() and then plotting. Use it to visualise class imbalance in datasets, frequency of categorical values, or the sample size per group before making any distributional comparison. Sort bars by frequency with order=df['col'].value_counts().index.

sns.pointplot() draws the mean for each category as a dot and connects them with a line, making it easy to see trends across ordered categories. Error bars show confidence intervals. This is particularly useful when the x-axis represents ordered groups (like age brackets or education levels) where you want to emphasise the direction and magnitude of change across groups, not just isolated means.

Test your understanding of Seaborn categorical plots from this lesson.

In this lesson you learned: sns.boxplot summarises distributions with median, IQR, and outliers, sns.barplot shows group means with confidence interval error bars, and sns.violinplot reveals the full distribution shape including multimodality. Next up we explore scatter plots and pair plots for visualising relationships between numeric variables.