Confidence Ellipse Plot
The 95% data ellipse of a bivariate sample, from the sample covariance and its eigenvalues.
What this page computes
Type the x and y values as comma-separated lists. The page takes the sample covariance matrix with the n − 1 divisor, finds its eigenvalues, and draws the ellipse whose semi-axes are √(χ²₂(0.95)·λ_i) with χ²₂(0.95) = 5.99146, rotated by half the angle whose tangent is 2r·s_x·s_y/(s_x² − s_y²). The points and the centre are drawn with it, and the semi-axes are printed above the plot.
This is a data ellipse describing spread, not a confidence region for the mean and not a 95% probability region for a future point; those need different scaling and the page does not mix them. A constant variable is refused, because the ellipse would have no area, and fewer than three pairs is refused.
Everything is computed here from the pasted pairs, so the same data always draws the same ellipse.
The page draws a confidence ellipse around a scatter of paired values, from the sample covariance matrix with the n − 1 divisor. The ellipse comes from the two eigenvalues and eigenvectors of that matrix, so its axes are the directions the data actually varies in rather than the x and y axes. Fewer than three pairs is refused, and so is a variable that never changes, because either leaves the ellipse with no area.
Draw a confidence ellipse over paired x and y values. Both lists are comma-separated, and the shape follows the sample covariance.
How to use it
Type the x values as a comma-separated list and the y values the same way; the two lists must be the same length, and a mismatch is refused with both counts named. The points are drawn with the ellipse over them, so a pair that sits outside the region is visible rather than described.
What the ellipse is made of
The covariances use the n − 1 divisor, which is the sample convention and matches the covariance page. Rotating the ellipse into its own axes turns the dependence between x and y into a tilt, which is the point of drawing it: a perfect circle would mean the two variables vary independently, and a long thin ellipse means one direction dominates. The page does not fit a model or test a hypothesis about the correlation.
What to Expect
Best for
- Seeing the joint spread of two variables at once
What this tool keeps
- Nothing leaves this tab: the arithmetic runs in the page you are reading, and a reload discards both the input and the result.
Frequently asked questions
How do I enter the two variables?
As two comma-separated lists of the same length — one list of x values and one of y values. Unequal lengths are refused with both counts shown.
Which divisor does the covariance use?
The sample convention, n − 1. That is the same divisor the covariance page prints first, so the ellipse and a quoted covariance agree.
Is the plot uploaded?
No. The covariance matrix, its eigen-decomposition and the SVG are all computed in this page, and a reload discards the plot.
Build custom tools for your store
Join the SDK program to create useful ecommerce utilities for your store.