Skip to contents

YAML Configuration for metaRVM

Source: vignettes/yaml-configuration.Rmd
yaml-configuration.Rmd

Introduction

The metaRVM package uses a YAML file to configure the model parameters. This vignette describes the structure of the YAML configuration file, starting with a simple example and progressively introducing more advanced features.

Basic Configuration

A minimal configuration file specifies the data sources, simulation settings, and disease parameters with fixed scalar values.

run_id: SimpleRun
population_data:
  mapping: data/demographic_mapping.csv
  initialization: data/population_init.csv
  vaccination: data/vaccination.csv
mixing_matrix:
  weekday_day: data/m_weekday_day.csv
  weekday_night: data/m_weekday_night.csv
  weekend_day: data/m_weekend_day.csv
  weekend_night: data/m_weekend_night.csv
disease_params:
  ts: 0.5
  tv: 0.25
  ve: 0.4
  dv: 180
  dp: 1
  de: 3
  da: 5
  ds: 6
  dh: 8
  dr: 180
  pea: 0.3
  psr: 0.95
  phr: 0.97
simulation_config:
  start_date: 01/01/2025 # m/d/Y
  length: 90
  nsim: 1
  random_seed: 42

Configuration Sections

  • run_id: A unique name for your simulation.
  • population_data: Paths to CSV files for population demographics, initial state, and vaccination schedules.
  • mixing_matrix: Paths to CSV files defining contact patterns for different times of the week.
  • disease_params: Disease characteristics. In this example, all parameters are single, fixed values.
  • simulation_config: Settings for the simulation run, such as start date, duration, and number of simulations.

Input File Structures

The metaRVM package requires several CSV files to be structured in a specific way. Below are the descriptions for each of the required input files.

Population Data Files

  • mapping: The population mapping file connects population IDs to demographic information. It must contain the following columns:
    • population_id: A unique identifier for each subpopulation, set of natural numbers 1, 2, 3, …
    • age: The age group of the subpopulation (e.g., “0-4”, “65+”).
    • race: The race or ethnicity of the subpopulation.
    • hcez: The healthcare zone or geographic region of the subpopulation.
  • initialization: This file specifies the initial state of the population for the simulation. It must contain the following columns:
    • N: The total number of individuals in each subpopulation.
    • S0: The initial number of susceptible individuals.
    • I0: The initial number of symptomatic infected individuals.
    • V0: The initial number of vaccinated individuals.
    • R0: The initial number of recovered individuals.
  • vaccination: The vaccination schedule file contains the number of vaccinations administered over time. The first column must be date in MM/DD/YYYY format, followed by columns for each subpopulation in the same order that they are assigned a population_id in the mapping file.

Mixing Matrix Files

The mixing matrix files define the contact patterns between different subpopulations. Each file should be a CSV without a header, where the rows and columns correspond to the subpopulations in the same order as the population mapping file. The values in the matrix represent the proportion of time that individuals from one subpopulation spend with individuals from another. The sum of each row must equal 1.

Disease Parameter Descriptions

Below is a list of the disease parameters used in metaRVM:

  • ts: Transmission rate for symptomatic individuals in the susceptible population.
  • tv: Transmission rate for symptomatic individuals in the vaccinated population.
  • ve: Vaccine effectiveness (proportion, range: [0, 1]).
  • dv: Mean duration (in days) in the vaccinated state before immunity wanes.
  • dp: Mean duration (in days) in the presymptomatic infectious state.
  • de: Mean duration (in days) in the exposed state.
  • da: Mean duration (in days) in the asymptomatic infectious state.
  • ds: Mean duration (in days) in the symptomatic infectious state.
  • dh: Mean duration (in days) in the hospitalized state.
  • dr: Mean duration (in days) of immunity in the recovered state.
  • pea: Proportion of exposed individuals who become asymptomatic (vs. presymptomatic) (range: 0-1).
  • psr: Proportion of symptomatic individuals who recover directly (vs. requiring hospitalization) (range: 0-1).
  • phr: Proportion of hospitalized individuals who recover (vs. die) (range: 0-1).

Defining Parameters with Distributions

Instead of fixed values, you can define disease parameters using statistical distributions. This is useful for capturing uncertainty in the parameters. metaRVM supports uniform and lognormal distributions.

Here is an example of defining ve, da, ds, and dh with distributions:

disease_params:
  ts: 0.7
  tv: 0.35
  ve:
    dist: uniform
    min: 0.29
    max: 0.53
  dv: 158
  dp: 1
  de: 3
  da:
    dist: uniform
    min: 3
    max: 7
  ds:
    dist: uniform
    min: 5
    max: 7
  dh:
    dist: lognormal
    mu: 8
    sd: 8.9
  dr: 187
  pea: 0.333
  psr: 0.95
  phr: 0.97
  • For a uniform distribution, you must specify min and max values.
  • For a lognormal distribution, you must specify mu and sd (mean and standard deviation on the log scale).

Specifying Subgroup Parameters

metaRVM allows you to specify different disease parameters for various demographic subgroups using the sub_disease_params section. These subgroup-specific parameters will override the global parameters defined in disease_params.

It is crucial that the demographic categories (e.g., age) and the specific values (e.g., 0-4, 5-11) used in this section exactly match the corresponding columns and values in the population mapping CSV file specified under population_data.

The following example defines different parameters for different age groups:

sub_disease_params:
    age:
      0-4:
        dh: 4
        pea: 0.08
        psr: 0.9303
        phr: 0.9920
      5-11:
        dh: 4
        pea: 0.08
        psr: 0.9726
        phr: 0.9920
      12-17:
        dh: 4
        pea: 0.08
        psr: 0.9726
        phr: 0.9920
      18-49:
        ts: 0.01
        dh: 6
        pea: 0.12
        psr: 0.9439
        phr: 0.9690
      50-64:
        dh: 6
        pea: 0.05
        psr: 0.9894
        phr: 0.9425
      65+:
        dh: 7
        pea: 0.05
        psr: 0.9091
        phr: 0.9227

In this configuration, individuals in the “0-4” age group will have a dh (duration of hospitalization) of 4, overriding any global dh value. Similarly, the transmission rate ts for the “18-49” group is set to 0.01.

Checkpointing and Restoring Simulations

For long-running simulations, it is useful to save the state of the model at intermediate points. This is known as checkpointing. metaRVM allows you to save checkpoints and restore a simulation from a saved state.

Enabling Checkpointing

To enable checkpointing, you need to add the checkpoint_dir and optionally checkpoint_dates to the simulation_config section of your YAML file.

  • checkpoint_dir: The directory where checkpoint files will be saved.
  • checkpoint_dates: A list of dates (in MM/DD/YYYY format) on which to save a checkpoint. If this is not provided, a single checkpoint will be saved at the end of the simulation.

Here is an example of how to configure checkpointing:

simulation_config:
  start_date: 01/01/2025
  length: 90
  nsim: 10
  random_seed: 42
  checkpoint_dir: "path/to/checkpoints"
  checkpoint_dates: ["01/15/2025", "01/30/2025"]

Restoring from a Checkpoint

To restore a simulation from a checkpoint file, use the restore_from parameter in the simulation_config section. This will initialize the model with the state saved in the specified checkpoint file.

simulation_config:
  start_date: 01/30/2025 # Should be the next date of the checkpoint date
  length: 60 # Remaining simulation length
  nsim: 10
  restore_from: "path/to/checkpoints/checkpoint_2025-01-30_instance_1.Rda"

When restoring, the start_date should correspond to the next date of the checkpoint, and the length should be the remaining duration of the simulation. Note that each instance of a simulation must be restored individually.