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
: Figurej
in the manuscript.figures/figj-v2.eps
: A variation of Figurej
. Here we modified the hypothesis prior distribution.figures/figj-v3.eps
: A variation of Figurej
. Here we modified the atoms prior distribution.figures/figj-vk.R
: Replication code forfigures/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 computefigures/figj-vk.eps
.data/figj-vk-f.csv
: Summary ofgamma
necessary to computefigures/figj-vk.eps
.data/figj-vk-f.jl
: Replication code forfigures/figj-vk-f.csv
.data/figj-vk-g.jl
: Replication code forfigures/figj-vk-g.csv
.
-
extras/
: contains auxilliary datasets and simulations.
To replicate all the results, do as follows:
- Install Julia and R.
- Instantiate the Pkg environment associated with this repository.
- Run each
.jl
file indata/
. - Run each
.R
file indata/
.
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), setNsim
to 100 indata/figj-vk-f.jl
anddata/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.