Skip to content

make_stackloss_check_data: Stack Loss Dataset

The make_stackloss_check_data function retrieves the classic Stack Loss dataset (Brownlee, 1965). This dataset describes the operation of a plant for the oxidation of ammonia to nitric acid and is a standard benchmark for robust regression due to the presence of well-known outliers.

Overview

The dataset consists of 21 operational days of a plant converting ammonia to nitric acid. The goal is typically to predict the Stack Loss (the amount of ammonia escaping unabsorbed) based on three operational variables.

  • Significance: Ideally suited for demonstrating robust regression methods because it contains several acknowledged outliers (observations 1, 2, 3, and 21) that can distort standard least-squares models.
  • Size: 21 samples, 4 variables.

Data Dictionary

The dataset is returned as a single matrix with the following columns:

  1. Air Flow (Feature): Rate of operation of the plant.
  2. Water Temp. (Feature): Cooling water inlet temperature.
  3. Acid Conc. (Feature): Acid concentration (in per 1000 minus 500).
  4. Stack.Loss (Target): Amount of ammonia escaping the absorption column.

Returns

Return Type Description
data numpy.ndarray The complete data array of shape (21, 4).
column_names list of str The list of columns: ['Air Flow', 'Water Temp.', 'Acid Conc.', 'Stack.Loss']

Example Usage

from machinegnostics.datasets import make_stackloss_check_data
import numpy as np

# Load the dataset
data, names = make_stackloss_check_data()

print(f"Data shape: {data.shape}")
print(f"Columns: {names}")

# Separate Features (X) and Target (y)
X = data[:, :3]
y = data[:, 3]

Source: Brownlee, K. A. (1965). Statistical Theory and Methodology in Science and Engineering. New York: John Wiley & Sons.

Author: Nirmal Parmar