Code Organization

[ShozenD]

Current Issues

• config.R file specifies too many paths / hard to understand / easy to go wrong
  • config.R file will have specification for confidential path files
    • Ollie: Disentangle confidential file paths from Normal file paths
    • Melodie: We can split the files. But people running the analysis shouldn’t worry too much about config.R contents
    • Ollie: Rarely the case that the user would just want to do a rerun
    • ACTION:
  • Ollie: Ideally, have config.yaml for experiment controls.
    • If usr wants to do sensitivity analysis, usr only needs to specify file path to new config
    • Melodie: The flags are changes automatically based on the job ID
      • transmission_flows/run_stan.R line 100 - 101
    • Ollie: Too convoluted. Maybe prepare 20 config files to specify parameters?
    • Melodie: If you were to change something between two runs, you would need until the job is running (HPC conflict)
    • Ollie: Have a config file for each job
  • Path to data is specified in the header

Anticipated User

• First time users

Anticipated Use Cases

• Advanced Case

  • User provides:
    • Data:
      • Incidence rates (per person/per year) (mean, upper 95% CI bound, lower 95% CI bound)
      • The number of unsuppressed by 1 year of age, gender
      • The number of susceptible by 1 year of age, gender
      • Counts by 1 year of age, gender of source recipient pairs
      • The contact rates by 1 year of age, gender
    • Model:
      • Advanced version for HPC (AS IS)
  • Outputs:
    • Convergence diagnostics
    • Marginal flows (age distribution of the sources Figure 2A)
    • Estimates need to create Figure 2B
    • One that investigates the model fit
      • How well does the marginalized estimates correspond to the prior
    • Figure 5C the outputs of the model against the prior

• Simple Case (Put on hold for now)

  • User provides:
    • Data:
      • Incidence rates (per person/per year) (mean, upper 95% CI bound, lower 95% CI bound)
      • The number of unsuppressed by 1 year of age, gender
      • The number of susceptible by 1 year of age, gender
      • Counts by 1 year of age, gender of source recipient pairs
      • The contact rates by 1 year of age, gender
    • Model:
      • Simple version? (run-able on a laptop)

NEEDS TO BE DONE

• Specify the format of the 5 data files that the users provide
• Tell the user how to feed input data to the model (run the analysis using original data)
• Create new scripts that provide these 5 data files for the Rakai analysis
• Remove from post-processing Figures and Tables for paper
• Version the current directory
• (Melodie) Remove Stan files from the previous analysis
• Tell the user how to make 5 input files

[ShozenD]

First release (DDL: Sept 20th)

• [x] Explain how the config.R file works in the first part of the README.md
• [x] Remove Stan files from the previous analysis (Andrea to communicate with Melodie)
• [ ] Write README for Zenodo (Ollie)

Second release

• [ ] Work on the simple use case

Notes

• Separate configurations and file paths by splitting config.R into two files, one containing paths and one containing configurations. (On hold)

[melodiemonod]

@ShozenD hi just to let you know that i have removed the deprecated stan files two months ago. Melodie