weightedpandas 0.1.0

Extends pandas Series and DataFrame classes to support weighted operations

WeightedPandas

WeightedPandas extends pandas Series and DataFrame classes to support weighted operations. It provides drop-in replacements for pandas objects that automatically handle weights in statistical operations.

Installation

pip install weightedpandas

Requirements

• Python 3.8+
• pandas 1.4.0+
• numpy 1.20.0+

Features

• Weighted versions of common statistical operations:
  • sum(), mean(), var(), std()
  • median(), quantile()
  • corr(), cov()
• Preserves weights through arithmetic operations
• Familiar pandas interface
• Supports both Series and DataFrame objects

Usage

import pandas as pd
import numpy as np
from weightedpandas import WeightedSeries, WeightedDataFrame

# Create a weighted series
data = [1, 2, 3, 4, 5]
weights = [5, 4, 3, 2, 1]
ws = WeightedSeries(data, weights=weights)

# Calculate weighted statistics
print(f"Weighted sum: {ws.sum()}")
print(f"Weighted mean: {ws.mean()}")
print(f"Weighted median: {ws.median()}")
print(f"Weighted standard deviation: {ws.std()}")

# Create a weighted dataframe
df_data = {
    'A': [1, 2, 3, 4, 5],
    'B': [5, 4, 3, 2, 1]
}
wdf = WeightedDataFrame(df_data, weights=weights)

# Calculate weighted statistics
print(wdf.sum())
print(wdf.mean())
print(wdf.corr())

# Weights are preserved through operations
ws2 = ws * 2 + 1
print(ws2.weights)  # Same as original weights

How Weights are Applied

In weighted calculations:

• sum(): Each value is multiplied by its weight before summing
• mean(): The weighted sum divided by the sum of weights
• var() and std(): Each squared deviation is weighted
• quantile(): The quantile is determined from the weighted cumulative distribution

Helper Functions

For convenience, you can use the following helper functions:

from weightedpandas import weighted_series, weighted_dataframe

# These are equivalent to the constructor calls
ws = weighted_series(data, weights=weights)
wdf = weighted_dataframe(df_data, weights=weights)

License

This project is licensed under the MIT License - see the LICENSE file for details.