Welcome to bayes_traj documentation!
bayes_traj is a Python package developed to perform Bayesian trajectory analysis, and it has several features that distinguish it from other trajectory approaches:
- It is fully Bayesian, so it supports incorporation of prior knowledge into the modeling process.
- It can simultaneously model multiple continuous and binary target variables as a functions of predictor variables.
- It uses Bayesian nonparametrics to identify the number of trajectories best supported by the data.
- Inference is performed using coordinate ascent variational inference, so it is fast and scales well to large data sets.
- Independantly estimates residual variance for each trajectory and each target variable.
- Can we used without random effects (group-based trajectory modeling, or GBTM), or with random effects (growth mixture modeling, or GMM).
- Provides a suite of tools to facilitate prior specification, model visualization, and summary statistic computation.
Installation
To get started using bayes_traj, simply install it using pip by running the following in your terminal:
pip install bayes_traj
bayes_traj provides several command-line tools:
generate_prior
-- used to speficy Bayesian priors for use the trajectory modelingviz_data_prior_draws
-- provides visualization of random draws from the priorbayes_traj_main
-- performs Bayesian trajectory modeling using a prior fileviz_model_trajs
-- provides visualization of trajectories fit usingbayes_traj_main
sumarize_traj_model
-- prints model summary and fit statistics given a model file produce bybayes_traj_main
assign_trajectory
-- writes a data file with appended trajectory assignment information given an input data file and a model file generated by thebayes_traj_main
tool
Each of these tools can be run with the -h flag for additional usage information.
Documentation Contents
- Tutorial -- provides a walk-through of basic usage
- Formulation -- states the Bayesian model assumptions and formulation
- Inference -- details the variational inference update equations