In yairgoldy/sl4hm: Simulated likelihood for modelling household epidemics

R/sl4hm: Simulated Likelihood for modelling household epidemics

Yair Goldberg and Rami Yaari

R/sl4hm is an R package. This package implements simulations of epidemics within households of children and adults and calculates the likelihood of a set of transmission parameters given a dataset of the result of epidemics (number of infected children and adults) within households.

Installation

You can install it from its GitHub repository. You first need to install the devtools package.

install.packages("devtools")

Then install R/sl4hm using the install_github function in the devtools package.

library(devtools)
install_github("yairgoldy/sl4hm")

The data

Loading the data used for the analysis and saving it to a csv file ca can be done as follows.

library(sl4hm)
household_data
write.csv(x = household_data,file = "household_data.csv")

The more detailed individualized-data level of the household data can be downloaded directly from this GitHub repository.

Example use

First create a dataset in which each row corresponds to a family and includes the number of adults and children and the index case.

family_structure_data <- create_family_structure_data(n=1000)
head(family_structure_data,n = 5)

Then simulate the epidemic dynamic within each household, given the transmission matrix for children and adults (beta) and a generation time distribution (P). The returned dataset includes the number of infected adults and children in each household.

dat <- create_simulated_data(family_structure_data = family_structure_data,
                             beta_cc = 0.1,
                             beta_ca = 0.2,
                             beta_ac = 0.2,
                             beta_aa = 0.4,
                             P=sl4hm::load_P())
head(dat,n = 5)

Calculate the likelihood surface for this dataset. While the simulated data can be generated with any combination of beta values, the likelihood function assumes a unique structure of beta in which beta_ac=beta_aadelta, beta_ca=beta_aagamma, and beta_cc=beta_aadeltagamma, where gamma and delta are non-negative coeficients measuring the susceptibility and infectivity of children compared to adults. The likelihood is calculated for a coarse grid. One can download a dictionary from here which enables calculating the likelihood on a finer grid. This dictionary holds the distribution of the number of infected for each family type and duration of epidemic. One can also create a dictionary using the function family_dictionary.

lik <- calc_likelihood(dat)
head(lik,5)
require(ggplot2,quietly = T)

ggplot(lik,aes(x=gamma_val,y=delta_val,z=nll))+
  facet_wrap(~beta_aa,labeller="label_both", ncol=3)+
  geom_contour_filled(binwidth=5)+
  geom_contour(color = "black", size = 0.1,binwidth=5) +
  xlab("Child/Adult susceptablity")+
  ylab("Child/Adult infectivity")+
ggtitle("Likelihood Surface")+
 theme_bw()

Using the function run_simulation one can create multiple datasets and find the MLE for each dataset. This can be used for calculating bootstrap confidence intervals for the parameters.

yairgoldy/sl4hm documentation built on Feb. 3, 2021, 5:45 p.m.