alabebm 0.4.0

Implementation of event-based models for degenerative diseases.

EBM

This is the python package for implementing Event Based Models for Disease Progression.

Installation

pip install alabebm

Change Log

• 2025-02-26. V 0.3.4. Modified the shuffle_order function to ensure full derangement, making convergence faster.

Generate Random Data

from alabebm import generate, get_params_path, get_biomarker_order_path
import os
import json 

# Get path to default parameters
params_file = get_params_path()

# Get path to biomarker_order
biomarker_order_json = get_biomarker_order_path()

with open(biomarker_order_json, 'r') as file:
    biomarker_order = json.load(file)

generate(
    biomarker_order = biomarker_order,
    real_theta_phi_file=params_file,  # Use default parameters
    js = [50, 100],
    rs = [0.1, 0.5],
    num_of_datasets_per_combination=2,
    output_dir='my_data',
    seed = None,
    prefix = None,
    suffix = None,
)

Run MCMC Algorithms

from alabebm import run_ebm
from alabebm.data import get_sample_data_path
import os

print("Current Working Directory:", os.getcwd())

for algorithm in ['soft_kmeans', 'conjugate_priors', 'hard_kmeans']:
    results = run_ebm(
        data_file=get_sample_data_path('25|50_10.csv'),  # Use the path helper
        algorithm=algorithm,
        n_iter=2000,
        n_shuffle=2,
        burn_in=1000,
        thinning=20,
    )

Input data

The input data should have four columns:

• participant: int
• biomarker: str
• measurement: float
• diseased: bool

An example is https://raw.githubusercontent.com/hongtaoh/alabEBM/refs/heads/main/alabEBM/tests/my_data/10%7C100_0.csv

The data should be in a tidy format, i.e.,

• Each variable is a column.
• Each observation is a row.
• Each type of observational unit is a table.

Features

• Multiple MCMC algorithms:

  • Conjugate Priors
  • Hard K-means
  • Soft K-means

• Data generation utilities

• Extensive logging