interpretablefa 6.1.0

A package for interpretable factor analysis

The interpretablefa package

Overview

This is a package, which is available on the Python Package Index, for performing interpretable factor analysis. This implements the priorimax rotation, including the associated interpretability index and plot, that are described here.

It also contains several helper and visualization functions and wraps several functions from the factor_analyzer package.

For more details about the methods implemented and some of the package's features, see Pairwise Target Rotation for Factor Models. The package's source code can be found here.

The linked paper provides some documentation to the package and docstrings are available. However, a comprehensive guide or documentation is still under development.

Example

For instance, suppose that data contains the dataset and q contains the questions. Then, one can fit a 4-factor priorimax model using the snippet below.

import pandas as pd
from interpretablefa import InterpretableFA

# Load the dataset and the questions
data = pd.read_csv("./data/ECR_data_clean.csv")
with open("./data/ECR_questions.txt") as questions_file:
    q = questions_file.read().split("\n")
    questions_file.close()

# Initialize the analyzer
analyzer = InterpretableFA(data)

# Fit the 4-factor model with the priorimax rotation
analyzer.fit_factor_model("model", 4, "priorimax", q)

# Get the results
print(analyzer.calculate_indices("model")["v_index"])
print(analyzer.models["model"].rotation_matrix_)

# Visualize the results
analyzer.var_factor_corr_plot("model")
analyzer.interp_plot("model")

Requirements

• Python 3.8 or higher
• numpy
• pandas
• scikit-learn
• factor_analyzer
• tensorflow_hub
• scipy
• seaborn
• matplotlib
• statsmodels

License

GNU General Public License 3.0