scagnostics

function scagnostics(ldIn, opts?): Scagnostics;

Defined in: scagnostics/index.ts:244

Nine measures of what the scatterplot looks like.

Shape descriptors, not quality measures: they say whether the plot is stringy, clumpy, outlier-ridden and so on, without reference to the data behind it. Use them to characterise a projection, or to compare projections’ visual character — not to decide which is faithful.

outlying, skewed, sparse, clumpy, striated, convex, skinny, stringy, monotonic.

• Needs: projection only, and it must be 2-dimensional. No labels.
• Range: each in [0, 1].
• Cost: O(N log N) — binning caps the work, so it is cheap even for large N.

Parameters

Parameter	Type
ldIn	PointsInput
opts	ScagnosticsOptions

Returns

Scagnostics

See

• Wilkinson, Anand & Grossman, IEEE InfoVis 2005 https://doi.org/10.1109/INFVIS.2005.1532142
• Wilkinson & Wills, J. Comput. Graph. Statistics 17 (2008) https://doi.org/10.1198/106186008X320465

Example

import { scagnostics } from "@saehrimnir/sickle";

// 2-D only, and it never looks at the high-dimensional data — these describe
// the shape of the scatterplot itself.
const s = scagnostics(projection);

s.clumpy;    // 0.9219 — clearly separated blobs
s.outlying;  // 0.1736
s.skinny;    // 0.5775