igutierrezm / manovabnpsupp

MANNOVABNPSupp

Replication code for all the figures and simulation results displayed in the manuscript "A new flexible Bayesian hypothesis test for multivariate data" (Gutiérrez et al. 2022). The repository is organized as follows:

• figures/: contains the figures and their associated replications code.

  • figures/figj-v1.eps: Figure j in the manuscript.
  • figures/figj-v2.eps: A variation of Figure j. Here we modified the hypothesis prior distribution.
  • figures/figj-v3.eps: A variation of Figure j. Here we modified the atoms prior distribution.
  • figures/figj-vk.R: Replication code for figures/figj-vk.eps.

• data/: contains the raw data as well as datasets with the posterior simulations.

  • data/figj-vk-f.csv: Summary of the density necessary to compute figures/figj-vk.eps.
  • data/figj-vk-f.csv: Summary of gamma necessary to compute figures/figj-vk.eps.
  • data/figj-vk-f.jl: Replication code for figures/figj-vk-f.csv.
  • data/figj-vk-g.jl: Replication code for figures/figj-vk-g.csv.

• extras/: contains auxilliary datasets and simulations.

To replicate all the results, do as follows:

1. Install Julia and R.
2. Instantiate the Pkg environment associated with this repository.
3. Run each .jl file in data/.
4. Run each .R file in data/.

Tip: We strongly recommend replicating our results using our Docker image. The necessary Dockerfile is in the root directory of this repository. We recommend this because our code has many non-trivial dependencies.

Tip: To reduce the computation time (and verify that everything is working correctly), the last 4 figures are computed using 1 simulated sample (instead of 100). If you want to use 100 samples in figures/figj-vk.eps (as we do in the manuscript), set Nsim to 100 in data/figj-vk-f.jl and data/figj-vk-g.jl and re-run the relevant scripts. Be aware that, depending on your computer, this can take several days to be done.