In kmcconeghy/flumodelr: An R Package for Estimating Attributable Influenza Morbidity and Mortality

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)
options(tibble.print_min = 4L, tibble.print_max = 4L)
library(flumodelr)
library(lubridate)
library(scales)

Create simulated dataset

First step, create a dataset, five years of influenza data in CDC format/canon.

#some have 53 weeks, but ignored for simulated dataset  
year <- rep(seq(2010, 2014, by=1), 52) 
week <- rep(seq(1, 52, by=1), 5) 

flu_sim <- tibble(year, week)
flu_sim <- mutate(flu_sim, 
                  week_in_order = row_number(),
                  season = if_else(week>=40 | week<20, T, F))

Data generating process

Total deaths

Total deaths obtained from actual CDC reported data for 2010-2015. It is assumed this value is accurate, but what is unknown are the proportion of deaths due to influenza.

flu_real <- fludta %>%
  dplyr::filter(week<=52)

ggplot(flu_real, aes(x=week_in_order, y=alldeaths)) +
  geom_line()

We add the real all-cause deaths to the dataset.

flu_sim$deaths_all <- flu_real$alldeaths

The objective is to simulate deaths attributable to influenza under different assumptions.

A seasonal influenza data generating process.
Features of influenza deaths: 1) Mild or severe season, qualitatively increases number of deaths. 2) Peak month, which month has peak
3) Peakedness, whether a sharp rise/fall, or gentle slope.

Peakedness simulated with vertex parabola

$$Eq \ 1. \ y = a(x-h)^2 + k$$
Where k is the highest observed influenza death, h is the peak week, and a determines peakedness.

#y = a(x-h)^2 + k

#peak of february, week 4
flu_sim_1 <- flu_sim %>%
  mutate(peak_week = row_number() - 4, 
         peak = dgamma(week, shape = 4.2, scale=1.2),
         deaths_flu_t = round(deaths_all*peak * season))

kmcconeghy/flumodelr documentation built on June 7, 2019, 8:47 p.m.