Detect outlier values using IQR/quantile method#

assesspy.is_outlier(x: list[int] | list[float] | Series, method: str = 'iqr', probs: tuple[float, float] = (0.05, 0.95), mult: float = 3.0) → Series#

Detect outliers in numeric values using standard methods.

Certain assessment performance statistics are sensitive to extreme outliers. As such, it is often necessary to remove outliers before performing a sales ratio study.

The IAAO standard method is to remove outliers that are 3 * IQR. Warnings are thrown when sample size is extremely small or when the IQR is extremely narrow. See IAAO Standard on Ratio Studies Appendix B. Outlier Trimming Guidelines for more information.

Parameters:

  • x (Array-like numeric values) – A list or pd.Series of numeric values, typically sales ratios. Must be longer than 2 and cannot contain Inf or NaN values.

  • method (str) – Default iqr. String indicating outlier detection method. Options are iqr or quantile.

  • probs (tuple[float]) – Upper and lower percentiles boundaries for the quantile method.

  • mult (float) – Default 3. Multiple of IQR to use as the outlier detection threshold.

Returns:

A boolean pd.Series the same length as x indicating whether or not each value of x is an outlier.

Return type:

pd.Series

Example:

# Detect outliers:
import assesspy as ap

ap.is_outlier(ap.ccao_sample().estimate)