Apply delay distribution to count or linelist data — sim_apply

Events include symptom onset, admission, death, test sampling, test processing

Usage

sim_apply_delay(
  df,
  ...,
  fn_p_symptomatic = ~0.5,
  fn_p_admitted = ~0.1,
  fn_p_died = ~0.05,
  fn_p_tested = ~0.8,
  seed = Sys.time()
)

Arguments

df

a line list dataframe arising from e.g. sim_branching_process()

...

Named arguments passed on to sim_apply_delay.linelist

fn_symptom_delay,fn_admission_delay,fn_death_delay: a function that calculates the time to event onset from infection. This will be called with a vector of infection times as the first parameter (time) but all other columns of df are also available as well as the symptomatic,died,and admitted flags. The function must be vectorised on its inputs (and consume additional inputs with ...). A purrr style lambda is OK e.g. ~ stats::rgamma(.x, shape = 3), and the first parameter will be infection time. if you have an discrete probability profile for this then you can use cfg_ip_sampler_rng(ip_symptoms).
fn_sample_delay: This function returns the time from either symptom onset (symptomatic) or from infection (asymptomatic) until a sample is taken. (N.B. this might be better to do a screening test probability plus screening test frequency rather than overloading this.)
fn_result_delay: Identical to other functions except the first parameter will be sample_time rather than time of infection. This is the time from sampling to the result being available.

Named arguments passed on to sim_apply_delay.count_data

fn_symptom_profile,fn_admission_profile,fn_death_profile: a function that takes time and returns the probability density of symptoms, admissions, or deaths over time since infection (i.e. tau) as an ip delay distribution. If possible it is a very good idea to pre-compute these distributions as they need to be assigned to every line in the input and this can be very slow.
fn_sample_profile: a function that takes time and returns the probability density of test sample being taken over time since symptoms.
fn_result_profile: a function that takes time and returns the probability density of test result being available over time since test sampling.

fn_p_symptomatic, fn_p_admitted, fn_p_died, fn_p_tested

Function that returns a probability between 0 and 1 for each row of the input dataframe. A purrr style lambda is OK (e.g. ~ 1 for always true) the first parameter of this will be time of infection. The function must be vectorised on its inputs (and consume additional inputs with ...)

seed

RNG seed for reproducibility

Value

Depends on input, either:

a wide format line list with additional XX, XX_time and XX_delay columns, for each of the set of statistics generated.
a long format set of counts of different statistics i.e. infections, symptoms, admission, death, sample (tests taken), results (test results) .

Examples

tmp = sim_branching_process(
  changes = tibble::tibble(t = c(0,20,40,60,80,110), R = c(1.8,1.5,0.9,1.5,0.8,1.2)),
  max_time = 120,
  seed = 100
)
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> .
#> complete

tmp2 = tmp %>% sim_apply_delay()
tmp2 %>% dplyr::glimpse()
#> Rows: 42,202
#> Columns: 20
#> $ time                <time_prd> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
#> $ id                  <int> 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,…
#> $ generation_interval <dbl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…
#> $ infector            <int> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA…
#> $ generation          <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,…
#> $ symptom             <lgl> FALSE, TRUE, TRUE, TRUE, FALSE, FALSE, TRUE, TRUE,…
#> $ symptom_delay       <dbl> NA, 1.4145076, 3.7457009, 1.5959996, NA, NA, 8.121…
#> $ symptom_time        <time_prd> NA, 1.4145076, 3.7457009, 1.5959996, NA, NA, …
#> $ admitted            <lgl> FALSE, FALSE, FALSE, TRUE, FALSE, FALSE, TRUE, FAL…
#> $ admitted_delay      <dbl> NA, NA, NA, 4.233242, NA, NA, 13.278138, NA, NA, N…
#> $ admitted_time       <time_prd> NA, NA, NA, 4.233242, NA, NA, 13.278138, NA, …
#> $ death               <lgl> FALSE, FALSE, FALSE, TRUE, FALSE, FALSE, FALSE, FA…
#> $ death_delay         <dbl> NA, NA, NA, 7.931225, NA, NA, NA, NA, NA, NA, NA, …
#> $ death_time          <time_prd> NA, NA, NA, 7.931225, NA, NA, NA, NA, NA, NA,…
#> $ tested              <lgl> TRUE, TRUE, TRUE, TRUE, TRUE, TRUE, TRUE, TRUE, TR…
#> $ sample_delay        <dbl> 2.3550527, 4.2549255, 0.3404628, 3.1655928, 1.6074…
#> $ sample_time         <time_prd> 2.3550527, 4.2549255, 0.3404628, 3.1655928, 1…
#> $ result              <lgl> TRUE, TRUE, TRUE, TRUE, TRUE, TRUE, TRUE, TRUE, TR…
#> $ result_delay        <dbl> 2.5151891, 4.6165771, 0.4029820, 3.4785471, 1.7402…
#> $ result_time         <time_prd> 4.8702418, 8.8715026, 0.7434448, 6.6441399, 3…