Exponential trace model for discrete and continuous data

This repository provides an implementation of the sampling-based approximation for computing the maximum likelihood estimator (MLE) of exponential trace model.

Usage

The file Implementation/exp_trace_model.R contains functions for computing the approximate maximum likelihood estimator. Our upcoming paper describes the algorithm in details.

We include an example code Implementation/example.R for analyzing neuron spike data. The neuron spike data is from Demas et al. 2003.

source("example.R")

Simulation

Running the simulation, as described in the paper, takes a long time and is recommended to be implemented on a cluster. We include small-scale example code in the repository.

Data-type specific functions can be found under Implementation/Data_type_specific.

Poisson analog data

source("Implementation/synthetic_Poisson_analog.R")

Exponential analog data

source("Implementation/synthetic_exponential_analog.R")

Composite of Poisson analog and Bernoulli data

source("Implementation/synthetic_Poisson_Bernoulli.R")

Composite of Poisson analog and Gaussian data

source("Implementation/synthetic_Poisson_Gaussian")

Repository Authors

• Rui Zhuang — Ph.D. candidate in Biostatistics, University of Washington — methodology and R implementation
• Noah Simon — Assistant Professor in Biostatistics, University of Washington — methodology
• Johannes Lederer — Professor in Mathematical Statistics, Ruhr-University Bochum — methodology