Distribution Plots: histplot and kdeplot

Seaborn is a statistical data visualisation library built on top of Matplotlib. It provides a high-level API that makes beautiful, informative plots with minimal code. One of the first things you want to understand about any dataset is the shape of its distributions — and Seaborn's histplot and kdeplot are the primary tools for this.

sns.histplot(data, x='column') creates a histogram of a numeric variable. Seaborn automatically chooses sensible bin counts using Sturges' rule by default. You can customise the number of bins with the bins parameter, or let Seaborn pick automatically. The height of each bar represents the count of observations in that range.

The bins parameter controls how many bars the histogram shows — more bins reveal finer detail but can look noisy. The stat parameter changes what the y-axis represents: 'count' (default), 'density' (probability density), 'probability' (fraction of observations), or 'percent'. Choosing stat='density' makes histograms from different-sized datasets comparable.

Setting kde=True in histplot overlays a Kernel Density Estimate (KDE) curve on the histogram. The KDE is a smooth continuous approximation of the underlying probability distribution. This combination is very effective: the histogram shows the raw bin counts while the KDE reveals the overall shape and detects multiple peaks (multimodality).

sns.kdeplot() draws only the smooth density curve without histogram bars. This is useful when comparing multiple distributions on the same axes because overlapping histograms become confusing, while overlapping KDE curves remain readable. The fill=True parameter shades the area under the curve, making it easier to distinguish groups visually.

Both histplot and kdeplot accept a hue parameter that splits the data by a categorical column and draws each group in a different colour. This makes it easy to compare distributions across groups. When using histplot with hue, set stat='density' to make groups of different sizes fairly comparable.

KDE has a tuning parameter called bandwidth that controls how smooth the curve is. A small bandwidth makes the curve jagged and overfits to the data; a large bandwidth over-smooths and hides real features. Seaborn uses bw_adjust (default 1.0) as a multiplier on the automatically chosen bandwidth — values below 1 increase roughness, above 1 increase smoothness.

Both histplot and kdeplot support two-dimensional plots by passing both x and y parameters. A 2D histogram shows a colour-coded frequency grid; a 2D KDE shows contour lines of equal density. These plots reveal the joint distribution of two numeric variables and whether they are correlated.

Seaborn provides built-in themes via sns.set_theme(style=). The available styles are 'darkgrid', 'whitegrid', 'dark', 'white', and 'ticks'. You can also set a palette with palette= or sns.set_palette(). These settings apply globally to all subsequent plots in the session, making it easy to achieve a consistent look.

When reading a distribution plot, look for four key features. Centre: where does the mass concentrate (mean/median)? Spread: how wide is the distribution (std dev)? Skewness: is the tail longer on the right (positive skew) or left (negative skew)? Modality: does the distribution have one peak (unimodal) or more (bimodal/multimodal)? Bimodal distributions often indicate two hidden subpopulations worth separating.

sns.displot() is the figure-level wrapper that creates its own Figure and supports facetting across rows and columns using col= and row= parameters. Pass kind='hist', kind='kde', or kind='ecdf' to switch plot types. The ECDF (Empirical Cumulative Distribution Function) is especially useful because it shows what fraction of the data lies below each value without requiring bin choices.

Test your understanding of Seaborn distribution plots from this lesson.

In this lesson you learned: sns.histplot creates histograms with customisable bins and stat parameters, sns.kdeplot draws smooth density curves ideal for group comparisons, and the hue parameter splits both plot types by a categorical variable for side-by-side comparison. Next up we explore categorical comparison plots like box plots, bar plots, and violin plots.