Detect Outliers

Overview

You send 1 through 100,000 finite numbers. IQR uses type-7 quartiles and iqrMultiplier 1.5 unless you override it in [0, 10]. ROBUST_Z uses median and MAD scaled by 1.482602218505602, with threshold 3.5 in (0, 20]. Constant data returns no outliers and does not divide by zero. Method-specific options on the wrong method are refused. This does not remove points or train a model.

A z-score that divides by a zero MAD invents outliers on a flat series. Fences and scaled MAD stay defined, and the constant case is empty.

When to use it

Scope

Non-goals

• Does not remove or impute the points
• Does not detect multivariate anomalies
• Does not train a model

Properties

Capabilities

Cost & billing

Flat rate per invocation

File handling

How we handle your data

Input schema

Output schema

API contract

curl -sS -X POST "$BASE/v1/execute" -H "Authorization: Bearer $KEY" -H 'Content-Type: application/json' -d '{"skillKey":"data.detect_outliers","input":{"values":[1,2,3,4,100]}}'

import { Squillify } from "@squillify/sdk";

const sq = new Squillify({ apiKey: process.env.SQUILLIFY_API_KEY });

const { output } = await sq.execute("data.detect_outliers", {"values":[1,2,3,4,100]});

Errors

Other limits

Limitations

• Does not remove or impute the points
• Does not detect multivariate anomalies
• Does not train a model
• Dropping the outliers in the same call
• Multivariate robust distance

How to call

POST /v1/execute (requires authentication)

{
  "skillKey": "data.detect_outliers",
  "input": {
    "values": [
      1,
      2,
      3,
      4,
      100
    ]
  }
}

Examples

{
  "values": [
    1,
    2,
    3,
    4,
    100
  ]
}

{
  "truncated": false,
  "upperBound": 7,
  "method": "IQR",
  "center": 3,
  "lowerBound": -1,
  "count": 5,
  "omittedCount": 0,
  "dispersion": 2,
  "outliers": [
    {
      "score": 46.5,
      "value": 100,
      "index": 4
    }
  ]
}

{
  "values": [
    5,
    5,
    5
  ],
  "method": "ROBUST_Z"
}

{
  "truncated": false,
  "upperBound": 5,
  "method": "ROBUST_Z",
  "center": 5,
  "lowerBound": 5,
  "count": 3,
  "omittedCount": 0,
  "dispersion": 0,
  "outliers": []
}

Recommended use cases

• Flag a 100 in an otherwise 1-through-10 series before a chart
• Confirm a constant series produces no divide-by-zero outliers

Avoid for

• Dropping the outliers in the same call
• Multivariate robust distance

Related skills

Composable with

Safety & privacy

• Runs without calling any external service.
• Your payloads are never stored permanently.
• Telemetry captures metadata only — never your payload contents.
• We never train models on your payloads.
• No human reviews your payloads.

Tags