History of Density 2D Plot
A density plot is a smoothed, continuous version of a histogram, estimated from data using kernel density estimation (KDE). In a 2d kernel density plot, a continuous curve called a kernel is drawn at each individual data point. These curves are then summed to produce a single smooth density estimation. The kernel most used is the Gaussian, which creates a bell-shaped curve at each point. This method forms the basis of the 2d density plot in data visualization, helping to reveal the underlying distribution more clearly than traditional histograms.
When to Use a 2D Density Plot?
With large sample sizes to reveal clear distribution patterns
Use a 2D density plot when data points are numerous and scatterplots risk overplotting. The 2D density plot counts observations within areas of the 2D space and uses colour gradients to highlight hidden density variations.
For nuanced density visualization
2D histograms and hexbins analyse relationships between two numeric variables by dividing the plot into polygons (squares or hexagons), reducing overplotting. A 2d kernel density estimation plot can also be represented with contours for smooth density interpretation.
To compare multiple distributions
Kernel density plots generally outperform histograms in visualizing several distributions simultaneously. Overlapping density curves allow easier separation of groups, whereas histograms often become cluttered. The continuous curves of kernel density estimation plots make density plot interpretation clearer and more effective in 2d data visualization.
Types of 2D Density Plots
1. Hexbin
Like a 2D histogram, but the plot area is divided into many hexagons instead of squares. This structure can better fit data and reduce visual bias.
2. 2D Histogram
A two-dimensional version of the classic histogram, the plot area is split into small squares, with each square’s colour representing the number of points it contains.
3. Contour Plot
A graphical method that represents a 3-dimensional surface on a 2D plane by plotting constant z-level slices called contours. In density visualization, contours represent areas of equal density, allowing clear density plot interpretation. Raster functions can also be applied to represent density as colours or shading.
When Not to Use 2D Density Plots?
When there aren’t enough data points to risk overplotting
If the dataset is small and scatterplots do not suffer from overplotting, use scatterplots instead. 2d density plots are most effective when data points overlap heavily, allowing colour gradients to represent density and reveal distribution patterns more clearly.
When you cannot adjust the plot’s bandwidth or bin size
The bandwidth or bin size for 2d density plots and histograms is critical and sensitive. If this parameter cannot be controlled or optimized for the data context, density plots may lead to misleading conclusions. In such cases, alternative plots should be used to represent density distributions more accurately.