Anomaly API

Flags the points that do not belong, marking them on the chart and noting them in the summary; for the full story see the Insights guide.

Try it - the outlier is flagged on the chart and called out in the summary:

anomaly

Import

import { anomaly, detectAnomalies } from "@michi-vz/insights/anomaly";

Signature & options

anomaly(options?) returns a plugin; its use() marks outliers on-chart and notes them in the summary. detectAnomalies(values: number[], options?) returns { method, anomalies, threshold }, where anomalies is an array of { index, value, score, kind }.

Both accept the same options:

Option	Type	Default	What it does
method	"zscore" or "iqr" or "forecast"	"zscore"	Detection strategy. "zscore" flags points far from the mean, "iqr" flags points outside the interquartile fences, and "forecast" flags points outside the prediction band.
threshold	number	~3 for zscore, ~1.5 for iqr	The z cutoff for "zscore" or the IQR k multiplier for "iqr". Optional.
target	string or string[]	all series	Restrict detection to these series. Optional.

Example

import { anomaly, detectAnomalies } from "@michi-vz/insights/anomaly";

// Standalone detection: flags the 50.
const result = detectAnomalies([10, 11, 9, 10, 50, 11]);
// result.anomalies -> [{ index: 4, value: 50, score, kind }]

// As a chart plugin, using the IQR method.
chart.use(anomaly({ method: "iqr" }));

Insights guide