A Python package for stepwise estimation of latent class models with measurement and structural components. The package can also be used to fit mixture models with various observed random variables.
Project description
StepMix
A Python package for latent class analysis and clustering of continuous and categorical data, with support for missing values. Various stepwise estimation methods are available for models with measurement and structural components. Largely based on Bakk & Kuha, 2018.
Install
You can install StepMix with pip, preferably in a virtual environment :
pip install stepmix
Tutorials
Detailed tutorials are available in notebooks :
- Latent Class Analysis with StepMix :
an in-depth look at how latent class models can be defined with StepMix. The tutorial uses the Iris Dataset as an example
and covers :
- Continuous LCA models;
- Binary LCA models;
- Categorical LCA models;
- Mixed LCA models (continuous and categorical data);
- Missing Values.
- Stepwise Estimation with StepMix :
a tutorial demonstrating how to define measurement and structural models. The tutorial discusses:
- LCA models with response variables;
- LCA models with covariates;
- 1-step, 2-step and 3-step estimation;
- Corrections and other options for 3-step estimation.
- Model Selection :
a short tutorial discussing:
- Selecting the number of latent classes (
n_components
); - Comparing models with AIC and BIC.
- Selecting the number of latent classes (
- Parameters, Bootstrapping and CI :
a tutorial discussing how to:
- Access StepMix parameters;
- Bootstrap StepMix estimators;
- Quickly plot confidence intervals.
Quickstart
A simple example for 3-step estimation on simulated data :
from stepmix.datasets import data_bakk_response
from stepmix.stepmix import StepMix
# Soft 3-step
X, Y, _ = data_bakk_response(n_samples=1000, sep_level=.9, random_state=42)
model = StepMix(n_components=3, n_steps=3, measurement='bernoulli', structural='gaussian_unit', assignment='soft',
random_state=42)
model.fit(X, Y)
print(model.score(X, Y)) # Average log-likelihood
# Equivalently, each step can be performed individually. See the code of the fit method for details.
model = StepMix(n_components=3, measurement='bernoulli', structural='gaussian_unit', random_state=42)
model.em(X) # Step 1
probs = model.predict_proba(X) # Step 2
model.m_step_structural(probs, Y) # Step 3
print(model.score(X, Y))
1-step and 2-step estimation are simply a matter of changing of the n_steps
argument. Additionally, some bias correction
methods are available for 3-step estimation.
References
-
Bolck, A., Croon, M., and Hagenaars, J. Estimating latent structure models with categorical variables: One-step versus three-step estimators. Political analysis, 12(1): 3–27, 2004.
-
Vermunt, J. K. Latent class modeling with covariates: Two improved three-step approaches. Political analysis, 18 (4):450–469, 2010.
-
Bakk, Z., Tekle, F. B., and Vermunt, J. K. Estimating the association between latent class membership and external variables using bias-adjusted three-step approaches. Sociological Methodology, 43(1):272–311, 2013.
-
Bakk, Z. and Kuha, J. Two-step estimation of models between latent classes and external variables. Psychometrika, 83(4):871–892, 2018
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.