SINC Algorithm to do simultaneous estimation of network of compisitional data and variable selection of additional covariates. The algorithm uses a novel Variational EM estimation scheme for computational efficiency. To increase compuation speed it is also built to support multiproccessor computing. A more detailed description of the model, algorithm, and applications can be found in

Osborne, N., Peterson, C.B. and Vannucci, M. (2020). Latent Network Estimation and Variable Selection for Compositional Data via Variational EM. Revised for Journal of Computational and Graphical Statistics.

All functions needed to run SINC are found in SINC_functions.py

SINC depends on numpy, scipy, multiprocess, and multiprocessing, and will need to be installed before using

This file contains the SINC function and all other dependent functions to run SINC. These are dependent on the python modules numpy, scipy, multiprocess, and multiprocessing.

A folder containing a small simulation tutorial on how the function may be used. In this folder you will find small simulation.ipynb, a notebook showing the models use, with supplementing synthetic data saved as csv files.

The simulations reported in section 4 of the paper can be found in this folder, labeled as Band, Cluster, Hub, and Random, corresponding to the graph structure.