pywinsor2 0.1.0

Python implementation of Stata's winsor2 command for winsorizing and trimming data

pywinsor2

Python implementation of Stata's winsor2 command for winsorizing and trimming data.

Installation

pip install pywinsor2

Quick Start

import pandas as pd
import pywinsor2 as pw2

# Load sample data
data = pd.DataFrame({
    'wage': [1.0, 1.5, 2.0, 3.0, 5.0, 8.0, 12.0, 20.0, 50.0, 100.0],
    'industry': ['A', 'A', 'B', 'B', 'A', 'A', 'B', 'B', 'A', 'B']
})

# Winsorize at 1st and 99th percentiles (default)
result = pw2.winsor2(data, ['wage'])

# Winsorize with custom cuts
result = pw2.winsor2(data, ['wage'], cuts=(5, 95))

# Trim instead of winsorize
result = pw2.winsor2(data, ['wage'], trim=True)

# Winsorize by group
result = pw2.winsor2(data, ['wage'], by='industry')

# Replace original variables
pw2.winsor2(data, ['wage'], replace=True)

Features

• Winsorizing: Replace extreme values with percentile values
• Trimming: Remove extreme values (set to NaN)
• Group-wise processing: Process data within groups
• Flexible percentiles: Specify custom cut-off percentiles
• Multiple variables: Process multiple columns simultaneously
• Stata compatibility: API designed to match Stata's winsor2 command

Main Function

winsor2(data, varlist, cuts=(1, 99), suffix=None, replace=False, trim=False, by=None, label=False)

Parameters:

• data (DataFrame): Input pandas DataFrame
• varlist (list): List of column names to process
• cuts (tuple): Percentiles for winsorizing/trimming (default: (1, 99))
• suffix (str): Suffix for new variables (default: '_w' for winsor, '_tr' for trim)
• replace (bool): Replace original variables (default: False)
• trim (bool): Trim instead of winsorize (default: False)
• by (str or list): Group variables for group-wise processing
• label (bool): Add descriptive labels to new columns (default: False)

Returns:

• DataFrame: Processed DataFrame with winsorized/trimmed variables

Examples

Basic Usage

import pandas as pd
import pywinsor2 as pw2

# Create sample data
data = pd.DataFrame({
    'wage': [1, 2, 3, 4, 5, 6, 7, 8, 9, 100],  # outlier: 100
    'age': [20, 25, 30, 35, 40, 45, 50, 55, 60, 25]
})

# Winsorize at default percentiles (1, 99)
result = pw2.winsor2(data, ['wage'])
print(result['wage_w'])  # New winsorized variable

# Winsorize multiple variables
result = pw2.winsor2(data, ['wage', 'age'], cuts=(5, 95))

# Trim outliers
result = pw2.winsor2(data, ['wage'], trim=True, cuts=(10, 90))
print(result['wage_tr'])  # Trimmed variable

Group-wise Processing

# Winsorize within groups
data = pd.DataFrame({
    'wage': [1, 2, 3, 10, 1, 2, 3, 15],
    'industry': ['A', 'A', 'A', 'A', 'B', 'B', 'B', 'B']
})

result = pw2.winsor2(data, ['wage'], by='industry', cuts=(25, 75))

Advanced Options

# Replace original variables
pw2.winsor2(data, ['wage'], replace=True, cuts=(2, 98))

# Custom suffix and labels
result = pw2.winsor2(data, ['wage'], suffix='_clean', label=True)

Comparison with Stata

Stata Command	Python Equivalent
winsor2 wage	pw2.winsor2(df, ['wage'])
winsor2 wage, cuts(5 95)	pw2.winsor2(df, ['wage'], cuts=(5, 95))
winsor2 wage, trim	pw2.winsor2(df, ['wage'], trim=True)
winsor2 wage, by(industry)	pw2.winsor2(df, ['wage'], by='industry')
winsor2 wage, replace	pw2.winsor2(df, ['wage'], replace=True)

License

MIT License

Author

Bryce Wang - [e-mail]

Contributing

Contributions are welcome! Please see CONTRIBUTING.md for guidelines.