In annecori/mRIIDSprocessData: Epidemiological forcasting using real-time data

knitr::opts_chunk$set(fig.width=12, fig.height=6, echo=FALSE, warning=FALSE, message=FALSE, fig.path = "output/figures/")

library(magrittr)
library(ggplot2)
library(dplyr)
library(EpiEstim)
devtools::load_all()

pow_dist    <- params$pow_dist
t.proj      <- params$t.proj
n.sim       <- params$n.sim
n.dates.sim <- params$n.dates.sim
p.stay      <- params$p.stay

Data clean-up

Read in data prepared using the above code chunks.

infile <- here::here("data", "CaseCounts/processed/WHO_bydistricts.csv")
WHO_bydistricts <- readr::read_csv(infile)

infile <- here::here("data", "CaseCounts/processed/WHO_bycountry.csv")
WHO_bycountry   <- readr::read_csv(infile)

Gravity model parameters

pow_N_to <- pow_N_from <- 1
K        <- 1

Comparison with the data from Health Map and ProMed

Projection using WHO data

Projection within Sierra Leone

Projection within Liberia

Projection within Guinea

Projection across all districts within Sierra Leone, Liberia and Guinea

Evaluating goodness of fit

outfile         <- "output/alldistricts-lh-tproj-100.csv"
log_likelihood  <- read.csv(outfile, na.strings = "-Inf")

## Replace -Inf by a large negative number

#log_likelihood[is.na(log_likelihood)] <- -900


pstay <- unique(log_likelihood$p.stay)   %>% sort
power <- unique(log_likelihood$pow_dist) %>% sort

z  <- select(log_likelihood, -c(pow_dist, p.stay)) %>%
      rowSums %>%
       matrix(nrow = length(power), byrow = TRUE)

p <- plotly::plot_ly(x = ~pstay, y = ~power, z = ~z) %>% plotly::add_surface()


#p <- plot_ly(x = ~pstay, y = ~power, showscale = TRUE) %>%
#  add_surface(z = ~z300) %>%
#  add_surface(z = ~z500, opacity = 0.98)