Skip to content

detect_outliers

Flag numeric cells that sit far from the rest of their column. Returns the coordinates of the flagged cells; it does not change the data.

When to use

  • Data-quality screening for suspicious numbers.
  • Locating extreme values before charting or aggregation.

Input schema

Parameter Type Required? Default Description
path string no* (no default) Path to the file (omit when open_tab is set)
open_tab string no (no default) Operate on an open GUI tab (@active or tab name)
table string no (no default) Specific table for multi-table sources
columns string[] no all columns Columns to check
method string no iqr iqr or zscore
k number no 1.5 IQR / 3.0 z Threshold multiplier

* path or open_tab is required.

IQR flags cells outside [Q1 - k*IQR, Q3 + k*IQR]. Z-score flags cells whose value is more than k standard deviations from the mean. Columns with fewer than four numeric values are skipped.

Response shape

{
  "flagged": [ { "row": 41, "column": "amount" }, … ],
  "count": 3,
  "method": "iqr",
  "k": 1.5
}

row is the zero-based row index; column is the column name.

Example call

{
  "name": "detect_outliers",
  "arguments": {
    "path": "/data/sales.parquet",
    "columns": ["amount"],
    "method": "zscore",
    "k": 3.0
  }
}

See also