Modeling Longitudinal Gambling Data: Challenges and Opportunities

This README consists of two parts: 1) the tutorial on how to fit the marginalized two part model, and 2) information about the simulation study.

Preprint: https://psyarxiv.com/uvxk2/ OSF link: https://osf.io/6pbgv/

Tutorial

The tutorial can be found in tutorial.html (HTML preview link), or the source document tutorial.Rmd. The tutorial shows how to fit a custom marginalized two-part gamma or lognormal model using brms.

Simulation

Requirements

The following folders must exist:

data/
save/
figures/

Configure the number of CPU cores and simulation replications in code/compute_setup.R

Required packages

install.packages(c("brms", 
                   "powerlmm", # version >= 0.5
                   "parallel",
                   "tidyverse", 
                   "ggstance", 
                   "cowplot"))

# Powerlmm dev version used for the simulations                   
devtools::install_local(path = "powerlmm_0.4.0.9000.tar.gz")

At the time of writing, powerlmm version 0.5.0 is not yet published on CRAN. A unfinished development version is therefore included with this repo.

Run the simulation

0_run_sims.R sources all the simulation files.

Results

These two files are used to summarize the results:

1_results.R Summarize all the results for the simulation with complete data, and saves the figures to figures\
1_results_MAR.R Summarizes MAR simulation and creates a table with the results

Additional code

The code used to simulate data from the hurdle model can be found in powerlmm:::sim_hurdle

rpsychologist / marginal-two-part