Visualising simulated data

In simulist: Simulate Disease Outbreak Line List and Contacts Data

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.width = 8,
  fig.height = 5
)

This vignette gives an overview of the ways to plot the line list and contacts data output from the sim_linelist() and sim_outbreak() functions.

Plotting can be useful to identify certain transmission dynamics and patterns in the simulated data, or just to check that the simulated data looks as expected given how the simulation was parameterised.

library(simulist)
library(epiparameter)
library(incidence2)
library(epicontacts)

First we load the required delay distributions using the {epiparameter} package.

contact_distribution <- epiparameter(
  disease = "COVID-19",
  epi_name = "contact distribution",
  prob_distribution = create_prob_distribution(
    prob_distribution = "pois",
    prob_distribution_params = c(mean = 3)
  )
)

infectious_period <- epiparameter(
  disease = "COVID-19",
  epi_name = "infectious period",
  prob_distribution = create_prob_distribution(
    prob_distribution = "gamma",
    prob_distribution_params = c(shape = 3, scale = 2)
  )
)

onset_to_hosp <- epiparameter(
  disease = "COVID-19",
  epi_name = "onset to hospitalisation",
  prob_distribution = create_prob_distribution(
    prob_distribution = "lnorm",
    prob_distribution_params = c(meanlog = 1, sdlog = 0.5)
  )
)

# get onset to death from {epiparameter} database
onset_to_death <- epiparameter_db(
  disease = "COVID-19",
  epi_name = "onset to death",
  single_epiparameter = TRUE
)

Setting the seed ensures we have the same output each time the vignette is rendered. When using {simulist}, setting the seed is not required unless you need to simulate the same line list multiple times.

set.seed(123)

Using a simple line list simulation with the factory default settings:

linelist <- sim_linelist(
  contact_distribution = contact_distribution,
  infectious_period = infectious_period,
  prob_infection = 0.33,
  onset_to_hosp = onset_to_hosp,
  onset_to_death = onset_to_death,
  outbreak_size = c(500, 1e4)
)

This line list contains r nrow(linelist) cases.

Visualising incidence of onset, hospitalisation and death

::: {.alert .alert-info}

This section of the vignette is heavily based upon examples given in the An introduction to {incidence2} package vignette. It is highly recommended to read the documentation supplied in the {incidence2} package to explore the full range of functionality.

:::

To visualise the number of cases with onset on a particular day, the {incidence2} package, and its dedicated class (<incidence2>) are used for handling and plotting this data.

Currently {simulist} outputs dates that are not rounded to the nearest day, i.e. it can be half way through a day. This is not obvious as R prints dates to the nearest day by default, and only by removing the date class (using unclass()) can you see the decimals (as R stores dates internally as the number of days since 1970-01-01).

Note storing dates as precise doubles and not as integer days may change in the near future.

The interval = "daily" is required as {incidence2} requires rounded dates to aggregate cases per unit time and specifying the interval will do this automatically for us. It is possible that not every date had the onset of symptoms, resulting in some dates missing entries. This is taken care of by setting complete_dates = TRUE, alternatively this can be achieved by using incidence2::complete_dates() on the <incidence2> object.

# create incidence object
daily <- incidence(
  x = linelist,
  date_index = "date_onset",
  interval = "daily",
  complete_dates = TRUE
)

#| fig.alt: >
#|   Histogram of incidence (epicurve) showing the daily incidence of cases
#|   from symptom onset including days with zero cases. The plot shows the
#|   outbreak increasing in incidence and then declining, similar to a normal
#|   distribution.
plot(daily)

Alternatively, incidence can be plotting weekly:

weekly <- incidence(linelist, date_index = "date_onset", interval = "isoweek")

#| fig.alt: >
#|   Histogram of incidence (epicurve) showing the weekly incidence of cases
#|   from symptom onset including days with zero cases. The plot shows the
#|   outbreak increasing in incidence and then declining, with some weeks late
#|   in the outbreak have noticably high case counts.
plot(weekly)

In order to check differences between a group in the line list data, for example sex, the <incidence2> data object can be recreated, specifying which columns to group by.

weekly <- incidence(
  linelist,
  date_index = "date_onset",
  interval = "isoweek",
  groups = "sex"
)

#| fig.alt: >
#|   Two histograms, in a single row, of incidence (epicurve) showing the
#|   weekly incidence of cases from symtom onset including days with zero
#|   cases, facetted by sex (left female, right male). The plot shows the
#|   outbreak increasing in incidence and then declining, with epidemic curves
#|   similar between sexes.
plot(weekly)

To visualise the onset, hospitalisation and death incidence in the same plot they can be jointly specified to the date_index argument of incidence2::incidence().

First the outcome data needs to be pivoted from long data to wide data to be input into incidence2::incidence().

Reshape line list data {.tabset}

Base R

# this can also be achieved with the reshape() function but the user interface
# for that function is complicated so here we just create the columns manually
linelist$date_death <- linelist$date_outcome
linelist$date_death[linelist$outcome == "recovered"] <- NA
linelist$date_recovery <- linelist$date_outcome
linelist$date_recovery[linelist$outcome == "died"] <- NA

Tidyverse

library(tidyr)
library(dplyr)
linelist <- linelist %>%
  tidyr::pivot_wider(
    names_from = outcome,
    values_from = date_outcome
  ) %>%
  dplyr::rename(
    date_death = died,
    date_recovery = recovered
  )

{-}

daily <- incidence(
  linelist,
  date_index = c(
    onset = "date_onset",
    hospitalisation = "date_admission",
    death = "date_death"
  ),
  interval = "daily",
  groups = "sex",
  complete_dates = TRUE
)

#| fig.alt: >
#|   Six histograms, in a grid of two columns and three rows, of incidence
#|   (epicurves) showing the daily incidence of cases from symtom onset
#|   (bottom row), hospital admission (middle row), and death (top row).
#|   The data is facetted by sex (left female, right male) and incidence event
#|   (rows). The plot shows the outbreak fluctuating, with many fewer
#|   hospital admissions and deaths than cases.
plot(daily)

Demographic data

Please see the Age-structured population vignette for examples of how to plot the distribution of ages within a line list data set, including age pyramids.

The plotting code in vignettes is hidden by default, click the Code button with arrow to reveal the plotting code.

Visualising contact data

::: {.alert .alert-info}

This section of the vignette is based upon examples from the {epicontacts} R package documentation and the examples provided in the The Epidemiological R Handbook chapter on transmission chains. We recommend going to the documentation of the {epicontacts} R package to see the all plotting and data wrangling functionality.

:::

Just as we utilised the <incidence2> class from the {incidence2} package to handle and plot incidence data, we are going to use the <epicontacts> class from the {epicontacts} R package to handle and plot epidemiological contact data.

::: {.alert .alert-success}

The benefit of using {epicontacts} is the same as {incidence2}, in the fact that a default plotting method is supplied by the package.

Advanced

Additionally, {epicontacts} provides access to network plotting from JavaScript libraries via the {visNetwork} and {threejs} R packages.

:::

The {epicontacts} function make_epicontacts() requires both the line list and contacts table, so we will run the sim_outbreak() function to produce both. We will use the same epidemiological delay distributions that we used to simulate a line list above, but reduce the mean number of contacts in the contact distribution to 2.

contact_distribution <- epiparameter(
  disease = "COVID-19",
  epi_name = "contact distribution",
  prob_distribution = create_prob_distribution(
    prob_distribution = "pois",
    prob_distribution_params = c(mean = 2)
  )
)

set.seed(1)
outbreak <- sim_outbreak(
  contact_distribution = contact_distribution,
  infectious_period = infectious_period,
  prob_infection = 0.5,
  onset_to_hosp = onset_to_hosp,
  onset_to_death = onset_to_death
)
head(outbreak$linelist)
head(outbreak$contacts)

Using the line list and contacts data simulated we can create the <epicontacts> object.

epicontacts <- make_epicontacts(
  linelist = outbreak$linelist,
  contacts = outbreak$contacts,
  id = "case_name",
  from = "from",
  to = "to",
  directed = TRUE
)

The <epicontacts> object comes with a custom printing feature to see the data.

epicontacts

To plot the contact network we can use the plotting method that is supplied by {epicontacts} and will be automatically recognised if the {epicontacts} package is loaded (as done above with library(epicontacts)).

::: {.alert .alert-primary}

If you are viewing this vignette on the web (or on a web browser) the graph below is interactive and will allow you to highlight individuals in the network using the drop-down menu, to zoom in and out of the plot by scrolling, and to move the network using the mouse to drag and drop.

:::

#| fig.alt: >
#|   Contact network from infectious disease outbreak. This includes all
#|   contacts, i.e. individuals that were infected and not infected. Each
#|   individual is shown as a circle with a name tag, and arrows connecting
#|   the contacts to show directionality of contact.
plot(epicontacts)

There is also the option to plot the contacts network in 3D using the epicontacts::graph3D().

By default the outbreak simulated by sim_outbreak() contains contacts of cases that were not infected. These are shown in the previous network plot by terminal nodes that do not pass on infection to other individuals (note that terminal nodes can also be infected individuals that did not infect anybody else, either due to not have any contacts or due to the probabilistic nature of infection transmission). Here we show how to subset the contacts table in order to only plot the transmission network of cases from the outbreak.

Subset contact network to transmission network {.tabset}

Base R

outbreak$contacts <- outbreak$contacts[outbreak$contacts$was_case == "Y", ]

Tidyverse

library(dplyr)
outbreak$contacts <- outbreak$contacts %>% # nolint one_call_pipe_linter
  dplyr::filter(was_case == "Y")

{-}

head(outbreak$linelist)
head(outbreak$contacts)

epicontacts <- make_epicontacts(
  linelist = outbreak$linelist,
  contacts = outbreak$contacts,
  id = "case_name",
  from = "from",
  to = "to",
  directed = TRUE
)

epicontacts

#| fig.alt: >
#|   Transmission chain from infectious disease outbreak. This includes only
#|   individuals that were infected, including confirmed, probable and
#|   suspected cases. Each individual is shown as a circle with a name tag,
#|   and arrows connecting the cases to show directionality of transmission.
plot(epicontacts)

Visualising other line list information

If there are other aspects of line list data that can be plotted and you would like to them added to this vignette please make an issue or pull request.

simulist documentation built on April 12, 2025, 1:34 a.m.