In reichlab/cdcfluutils: Utility functions for CDC flu forecasts for team Kernel of Truth

This vignette demonstrates the functionality provided by the cdcfluutils package for sampling possible revisions to currently observed weighted ILI based on historical revisions. This is in the rRevisedILI function written by Casey Gibson.

Sampling revisions to national level wILI data

library(dplyr)
library(ggplot2)
library(cdcfluutils)

data_nat <- nat_reg_flu_data_with_backfill %>% filter(region == "nat")
data_nat_subset <- data_nat[data_nat$epiweek %in% seq(201731,201750) & data_nat$issue == 201750,]$wili
n <- 1000

# sampling historical revision trajectories
sampled_trajs <- rRevisedILI_cached(n,data_nat_subset,50,"nat","2017/2018", season_start_epiweek = 31)

sampled_trajs_df <- data.frame(
  y = c(t(sampled_trajs)),
  x = rep(1:(length(data_nat_subset)), n),
  group = rep(1:n, each=length(data_nat_subset))
)

p_traj <- ggplot(data = sampled_trajs_df,
    mapping = aes(x = x, y = y, group = group)) +
  geom_line(alpha=.05) +
  theme_bw() +
  geom_line(data = data.frame(y = data_nat_subset, x = 1:length(data_nat_subset)),
    mapping = aes(x = x, y = y, col = 'Observed', group = 1)) +
  ggtitle("Sampling historical revision trajectories")

# sampling historical revision trajectories
sampled_trajs <- rRevisedILI_fast(n,data_nat_subset,50,"nat","2017/2018", season_start_epiweek = 31)

sampled_trajs_df <- data.frame(
  y = c(t(sampled_trajs)),
  x = rep(1:(length(data_nat_subset)), n),
  group = rep(1:n, each=length(data_nat_subset))
)

p_norm <- ggplot(data = sampled_trajs_df,
    mapping = aes(x = x, y = y, group = group)) +
  geom_line(alpha=.05) +
  theme_bw() +
  geom_line(data = data.frame(y = data_nat_subset, x = 1:length(data_nat_subset)),
    mapping = aes(x = x, y = y, col = 'Observed', group = 1)) +
  ggtitle("Sampling from normally distributed revisions")

library(gridExtra)
grid.arrange(p_traj, p_norm)

reichlab/cdcfluutils documentation built on March 12, 2020, 2:49 p.m.