This GitHub repository accompanies the paper by Kipshidze, Klein, and Yang titled "Understanding the drivers of continued mpox transmission in the United States: a modeling study." The code and data files will receive the simulations, analyses, and figures for this paper.
We recommend that these models be run on a remote server. We utilized an c7i.48xlarge instance on Amazon Web Serivvice WS) on a Linux platform with base R and associated packages. The average run-time for our model was approximately 4-7 minutes per stochastic run when using parallelization.
You will need to set up a shell script to run and pass arguments to the run_LHS sims.R script. After running the models you will then use post_estimating fit.R to compute the negative-log likelihood of each individual run, which is compared to the observed data. This script is set up that it can be pointed to the saved runs and do the rest for you. Finally, the post_assess fit.R will take the estimated NLL and filter down to the top 10% of all runs for each city and plot the fitted runs and distributions of the free parameters. Finally the script post_estim shape params.R will estimate the shape parameters.
You will need to set-up a shell script to pass the different arguments of applicable mechanisms, the arguments that can be passed are listed below in the table. These arguments would get passed to run_mechanisms.R
| Argument | Description |
|---|---|
| asymp_switch | Whether underdetection alone is enabled, logical T/F |
| reduce_asymp | Whether underdetection should be reduced by some proportion, logical T/Fic |
| reinfection | Whether reinfections should be enabled, logical T/F |
| reinfection_time | If reinfections are enabled, the timing in months. |
| vaxwane | Whether vaccine-waning is enabled, logical T/F. |
| vaxwane_p | The number of months to median time for waning, partial vaccination |
| vaxwane_f | The number of months to median time for waning, full vaccination |
| post_pride_seeds | Whether additional seeding should be enabled post-pride 2022. |
| mech_name | A name for the mechanism being evaluated. |
Each mechanism was run 1,000 times for each city. To improve efficiency, we sent these runs across 90+ nodes on AWS, meaning each node saves its own .Rdata file. We created several scripts that read all the files and collapse and summarize the runs into the final figures seen in our manuscript. All relevant scripts that deal with summarizing and plotting simulations start with the prefix post.
Kipshidze N., Klein E., Yang W. (2023). Understanding the drivers of continued mpox transmission in the United States: a modeling study. GitHub. https://github.com/CDDEP-DC/mpox-modeling-study