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:
  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
  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
  nrep: 1
  simulation_mode: deterministic
  random_seed: 42

Configuration Sections

  • run_id: A unique name for the 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, along with examples of what they should look like.

Population Data Files

  • initialization: This file specifies both the demographic structure and initial state of the population for the simulation. It must contain the following required columns:
    • population_id: A unique identifier for each subpopulation (sequential natural numbers: 1, 2, 3, …)
    • 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
    User-defined demographic category columns: Any additional columns beyond the required ones are automatically detected as demographic categories. These can be used for stratification and in sub_disease_params for category-specific parameters. Common examples include:
    • age: Age groups (e.g., “0-17”, “18-64”, “65+”)
    • race: Race or ethnicity categories
    • zone: Healthcare zones or geographic regions
    • Or any custom categories like income_level, occupation, risk_group, etc.

Example of a population initialization file:

#> First 10 rows of population_init_n24.csv:
#>    population_id   age race zone      N     S0  I0 V0 R0
#> 1              1  0-17    A   11  30742  30711  31  0  7
#> 2              2 18-64    A   11  41429  41388  41  0  9
#> 3              3   65+    A   11   3321   3318   3  0  0
#> 4              4  0-17    B   11  70138  70068  70  0  6
#> 5              5 18-64    B   11  12298  12286  12  0 32
#> 6              6   65+    B   11  11549  11537  12  0  0
#> 7              7  0-17    C   11  84178  84094  84  0 13
#> 8              8 18-64    C   11 113521 113407 114  0 15
#> 9              9   65+    C   11  11924  11912  12  0  0
#> 10            10  0-17    D   11 199686 199486 200  0 14
#> 
#> ... (24 total rows)
  • 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 as population_id in the initialization file.

Example of a vaccination schedule file:

#> First 10 rows of vaccination_n24.csv:
#>          date v1 v2 v3  v4 v5 v6  v7  v8  v9 v10 v11 v12 v13 v14 v15  v16 v17
#> 1  09/16/2023  0  0  0   0  0  0  11   5   4   8   6   4   2   5   1    8  11
#> 2  09/23/2023  5 12  1  11  7  2 180 187  99 362 293  72 311 163  24  519 203
#> 3  09/30/2023  3  5  2  19 10  3 198 235  88 424 364  89 445 255  36  625 251
#> 4  10/07/2023 20 14 15  46 19  8 403 385 124 846 580 187 675 389  45 1242 347
#> 5  10/14/2023 28 29 19  81 50  7 471 425 105 997 511 162 727 400  42 1351 313
#> 6  10/21/2023 37 55 24 110 43  8 378 499 115 980 483 160 623 511  39 1212 317
#> 7  10/28/2023 28 43 19  86 48 16 339 464  88 809 438 126 548 411  52 1090 287
#> 8  11/04/2023 22 53 19  91 49 28 329 391  70 765 431 147 522 391  22  956 259
#> 9  11/11/2023 36 62 21 110 60 10 337 396  87 769 414 148 575 425  39  966 199
#> 10 11/18/2023 40 65 14 102 55 14 340 431  97 875 392 118 513 417  28 1041 231
#>    v18 v19 v20 v21 v22 v23 v24
#> 1    5   0   0   0   0   0   3
#> 2  113   2   3  26   4  13  10
#> 3  119   4  22  27   2  18  27
#> 4  193   8  41  50   4  70  44
#> 5  161  10  98 217  36 123  98
#> 6  202  18 160 214  51 149 168
#> 7  152  32 199 273  60 153 267
#> 8  126  26 168 239  33 149 225
#> 9  149  26 197 255  52 142 275
#> 10 141  26 212 242  50 153 191
#> 
#> ... (51 total rows)
#> 
#> Note: Columns represent vaccination counts for each population_id (1-24)

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 population_id in the initialization 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.

Example of a mixing matrix file (weekday day):

#> First 10 rows and 10 columns of m_weekday_day.csv:
#>             V1           V2          V3          V4           V5          V6
#> 1  0.860000000 0.0010475811 0.003377249 0.001710217 0.0011920273 0.006571116
#> 2  0.003611954 0.9100000000 0.001328864 0.003596329 0.0027309903 0.004173784
#> 3  0.001344397 0.0072452422 0.870000000 0.002513688 0.0050834111 0.009789632
#> 4  0.006689671 0.0005482346 0.004749909 0.920000000 0.0070678313 0.001806128
#> 5  0.007740689 0.0039632387 0.004643154 0.001091787 0.8700000000 0.009791262
#> 6  0.006379144 0.0032865609 0.003432134 0.001761560 0.0071883914 0.890000000
#> 7  0.001444749 0.0064237517 0.002656119 0.001968115 0.0043013059 0.001206245
#> 8  0.002179355 0.0070411463 0.001587234 0.003439760 0.0033650992 0.006414028
#> 9  0.005390480 0.0066530217 0.005482305 0.003007641 0.0006105881 0.003145261
#> 10 0.001666809 0.0029011017 0.008534104 0.001698315 0.0098720801 0.004586367
#>              V7           V8           V9          V10
#> 1  1.464237e-02 0.0143653234 0.0118876071 0.0080742019
#> 2  6.528305e-03 0.0041375710 0.0057636782 0.0054420004
#> 3  8.763376e-03 0.0005057071 0.0007488488 0.0073198670
#> 4  2.822754e-04 0.0070548564 0.0009509153 0.0012611329
#> 5  1.004983e-02 0.0041073157 0.0088236656 0.0006445009
#> 6  3.144937e-05 0.0085528740 0.0063335264 0.0023219191
#> 7  9.200000e-01 0.0035656434 0.0070237063 0.0042256901
#> 8  4.098307e-03 0.9100000000 0.0067478716 0.0100518811
#> 9  9.597300e-04 0.0002268853 0.9200000000 0.0009418937
#> 10 5.236045e-03 0.0065183293 0.0012107981 0.8700000000
#> 
#> Matrix dimensions: 24 x 24
#> Row sums (should all equal 1):
#>  [1] 1 1 1 1 1 1 1 1 1 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.
  • 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, disease parameters can be defined 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: 0.5
  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, min and max values must be specified.
  • For a lognormal distribution, mu and sd (mean and standard deviation on the log scale) must be specified.

Specifying Subgroup Parameters

metaRVM allows different disease parameters to be specified for demographic subgroups using the sub_disease_params section. These subgroup-specific parameters override the global parameters defined in disease_params.

The demographic categories used in this section must match the user-defined category column names in the initialization CSV file specified under population_data. For example, if the initialization file has columns named age, income_level, and occupation, any of these categories can be used in sub_disease_params. The specific values (e.g., "0-4", "low", "healthcare") must exactly match the values in those columns.

The following example defines different parameters for different age groups:

sub_disease_params:
    age:
      0-17:
        dh: 4
        pea: 0.08
        psr: 0.9303
        phr: 0.9920
      18-64:
        dh: 4
        pea: 0.08
        psr: 0.9726
        phr: 0.9920
      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.

Stochastic Simulation with Distributional Parameters

When both parameter uncertainty and stochastic disease transitions are represented, set simulation_mode: stochastic and define one or more disease parameters as distributions.

  • nsim: number of sampled parameter sets
  • nrep: number of stochastic replicates per parameter set
  • total runs = nsim * nrep

Example:

run_id: StochasticDistRun
population_data:
  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
  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: 2.0
    sd: 0.5
  dr: 187
  pea: 0.333
  psr: 0.95
  phr: 0.97
simulation_config:
  start_date: 01/01/2025
  length: 90
  nsim: 20
  nrep: 5
  simulation_mode: stochastic
  random_seed: 42

For reproducibility, provide random_seed. This seed is used to reproduce both the parameter draws (for distributional parameters) and the stochastic model replicates.

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 checkpoints to be saved and simulations to be restored from a saved state.

Enabling Checkpointing

To enable checkpointing, checkpoint_dir and optionally checkpoint_dates need to be added to the simulation_config section of the 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
  nrep: 1
  simulation_mode: deterministic
  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, the restore_from parameter is used in the simulation_config section. The model is initialized 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
  nrep: 1
  simulation_mode: deterministic
  random_seed: 42
  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.