meanvar-clustering-longitudinal-data

Supplementary materials for the manuscript "Latent-class trajectory modeling with a heterogeneous mean-variance relation" by N. G. P. Den Teuling, F. Ungolo, S.C. Pauws, and E.R. van den Heuvel

This repository contains the source code for the conditional growth mixture Stan models, the estimation of the marginal loglikelihood thereof, and for running and analyzing the simulation study and case study.

Setup

1. Install R.
2. Create an .Rprofile file with the following content, and fill in the placeholders:

source("renv/activate.R")
FIG_DIR = 'figs'
RESULTS_DIR = 'results'
TABLES_DIR = 'tables'
COVID_DATA_DIR = '~/data/csse_covid_19_data' # set to correct folder
REDIS_HOST_FILE = file.path('redis', 'redis_host.txt') # used by worker.R to connect to the Redis server
options(
  redis.host = 'localhost',
  redis.port = 6379,
  redis.pwd = '', # set password if configured
  latrend.warnMetricOverride = FALSE,
  mc.cores = parallel::detectCores(logical = FALSE)
)

source('include.R')

3. Install the required packages via renv::restore() or manually.

Setting up and using the batch job environment

In case you want to run the simulation or case study, proceed with the next steps.

1. Install Redis.
2. Configure Redis and the credentials in the .Rprofile file.
3. Run Redis
4. Test if you can connect from R by running sim_init().
5. Submit jobs, e.g., by running sim_all.R.
6. Run worker.R as one or more stand-alone processes, e.g., by executing worker6.bat on Windows.
7. Wait a long time for computations to finish.
8. Collect and process results (in case of the simulation study) by running process_results.R.

Case study results