XY Plot

Overview

The XY plot maps two numeric columns against each other — one as the X value, one as the Y value. It is the standard tool for making relationships, trends, and patterns between two variables visible.

Dataset: A dataset is one X/Y combination drawn as a series of points. Each dataset has its own X and Y column — datasets are not required to share the same X axis.

Markers and lines: Markers show individual measurements. Lines connect the points in data order — useful for time series or ordered traces, misleading for unordered samples.

Reference lines and bands: Optional horizontal or vertical reference lines (e.g. target value, specification limit) and colored bands (tolerance band, acceptable range) help compare measurements against an expected corridor.

Unlike the scatter plot in the correlation module, the XY plot is purely descriptive — it does not run any statistical test and is used for visual exploration.

Operation

• Add a dataset: pick an X column and a Y column — the plot updates immediately.
• Multiple datasets: use "Add dataset" to lay down more X/Y pairs for group comparisons.
• Markers or lines: global toggles "Show lines" and "Show markers" control the appearance of all datasets.
• Per dataset, color, symbol, line width, and name can be customized in the editor.
• Reference lines and bands can be added in the editor panel — ideal for specification limits, target values, or trend lines.

The chart can be exported as PNG or SVG and follows the active theme (light/dark).

What to look for

• Points line up along a rising line → positive relationship between X and Y.
• Points line up along a falling line → negative relationship.
• Points scatter with no direction → no relationship or pure random noise.
• Arc or U-shape → nonlinear relationship (linearize for correlation or use Spearman).
• Funnel-shaped scatter → heteroscedasticity (spread grows with X) — important for regression.
• Several separate clouds → mixed populations, stratify the data.

For a quantitative statement on the strength and significance of the relationship, follow up with the correlation or regression module.

Pitfalls

Lines on unordered data: Lines connect points in row order — for unordered samples this produces a meaningless zig-zag. Only enable lines when the X axis carries a real order (time, index, position).

Visual relationship ≠ correlation: Even a clearly visible pattern does not automatically mean a statistically significant relationship. When in doubt, follow up with the correlation module and check Pearson or Spearman coefficients with p-values.

Axis scaling distorts the picture: Too narrow or too wide a scale can make strong trends look harmless or harmless scatter look dramatic. Choose axes deliberately — the default scaling is usually, but not always, sensible.

Outliers dominate the picture: A single extreme value shifts the axis scaling and squeezes the remaining points together. Mark outliers, investigate them, and consider excluding them in a separate view.

Overplotting: With many data points, markers overlap and the true density becomes invisible. Shrink markers, draw them more transparently, or look at a histogram or boxplot in addition.

Bubble plot (third column as size)

When the S-slot (Size) of a dataset is also filled, the result is a bubble plot: marker size encodes a third numeric column, allowing three variables to be shown at once.

Area-proportional scaling: Raw values are mapped through a square-root normalisation to a 6–28 px diameter range. As a result the bubble AREA (not the radius) is proportional to the value — visually correct according to perceptual research.

Negative or missing values: Rows with missing, non-numeric, or negative size values are silently skipped — a negative area has no sensible visual representation.

• Typical uses: project prioritisation (effort × benefit × risk), marketing (CPC × conversion × reach), machines (cycle time × scrap × output).
• The tooltip shows the raw value of the size column in addition to X and Y.
• Combined with the grouping slot (G): each group receives its own bubble normalisation within the group.

Pitfalls: when all size values are nearly equal, differences disappear visually (sqrt scaling damps them further). When the range is very wide, mid-range values become too small. In both cases transform the column up front or filter it.

Examples

This module ships with the following example datasets — load any of them in the app with a single click.