marinavannucci / slam

R code for Monte Carlo EM algorithm of Semiparametric Latent ANOVA Model (SLAM)

SLAM

The slam package is used for implementing the Monte Carlo EM algorithm proposed in the paper ``Semiparametric Latent ANOVA Model for Event-Related Potentials". The algorithm estimates the latency of ERP signals with the full posterior distribution at both subject level and group level.

Installation and Usage

You can install slam from GitHub with:

# install.packages("devtools")
devtools::install_github("chenghanyustats/slam")

Some important functions are

• mcem_slam() implement the proposed MCEM algorithm for SLAM.

• Rsolnp::solnp() is used to optimize the hyperparameters of the kernel function used in the GPR.

• get_pi_t_sig() does the regression curve fitting and computes the associated credible interval.

• plot_pred_gp_f_y() plots the fitted results from get_pi_t_sig().

• get_amp_*() family estimates the amplitude of the ERP components.

Package Dependency

The required R packages include

packages <- c("Matrix", "matrixcalc", "emulator", "doParallel", "Rsolnp", 
          "mvnfast", "extraDistr", "truncnorm", "invgamma", "ggridges",
          "ggplot2", "tidyverse")

The R version >= 3.5 is required.

Vignette

The package provide a vignette slam.Rmd(slam.pdf) for implementing the proposed method including simulation, and ERP case studies. The code reproduces some key figures shown in the paper.