Chapter 4 DENGUE DATA

4.1 Init

library(tidyverse)
library(lubridate)
library(rEDM)
library(ISOweek)
library(stringr)
library(cowplot)
library(purrr)
library(naniar)
library(ggsci)

4.2 Data

rm(list = ls())

load("./_data/climate_vars/Data.RData")

den <- read_csv("_data/LOR_dengue.csv")

nl <- nl %>%
  select(-ntl2014)

nl2 <- readRDS("./_data/climate_vars/NightTimeLight.rds")

ubi_den <- unique(den$ubigeo)
ubi_clim <- unique(temp$UBIGEO) %>% as.numeric()

unique(ubi_clim[! ubi_clim %in% ubi_den]) # Rosa Panduro creado 2014

## [1] 160802

4.3 Assembly Climate

data_frames <- do.call("list",mget(ls())) %>%
  enframe() %>%
  #filter(name == "nl2") %>%
  filter(!name %in% c("den", "ubi_clim","ubi_den", "data_frames")) %>%
  mutate(dat_edit = map(.x = value, 
                       .f = ~gather(.x, var_date, value, 5:last_col()) %>%
                         separate(var_date, into = c("var", "date"), 
                                  sep = "(?<=[A-Za-z])(?=[_0-9])") %>%
                         mutate(date = str_sub(date, -7,-1)) %>%
                         separate(date, into = c("year", "month"),
                                  sep = "-", convert = T) %>%
                         mutate(month = ifelse(is.na(month), 1, month)))) %>%
  select(name, dat_edit) %>%
  unnest(cols = "dat_edit") %>%
  select(-name) %>%
  spread(var, value) %>%
  filter(year<2021) %>%
  mutate(UBIGEO = as.numeric(UBIGEO))
  # 53 districts * 12 months * 21 years = 13,356

# a <- nl2 %>%
#   gather(var_date, value, 5:last_col()) %>%
#   separate(var_date,
#            into = c("var", "date"),
#            sep = "(?<=[A-Za-z])(?=[_0-9])"
#            ) %>%
#   mutate(date = str_sub(date, -7,-1)) %>%
#   separate(date,
#            into = c("year", "month"),
#            sep = "-",
#            convert = T) %>%
#   mutate(month = ifelse(is.na(month), 1, month))

vis_miss(data_frames %>%
           arrange(year, month))

4.4 Assembly Cases

den_proto <- den %>%
  filter(complete.cases(.),
         yy<2021) %>%
  expand(ubigeo, yy, mm) %>% # 52 districts * 12 months * 21 years = 13,104
  full_join(den %>%
              select(ubigeo, yy, mm, tot_casos) %>%
              filter(yy<2021), by = c("ubigeo", "yy", "mm")) %>%
  mutate(tot_casos = replace_na(tot_casos, 0)) %>%
  full_join(den %>%
              distinct(ubigeo, departam, provincia, distrito), by = "ubigeo") %>%
  rename(UBIGEO = ubigeo,
         year = yy,
         month = mm,
         department = departam,
         province = provincia,
         district = distrito,
         den_cases = tot_casos)

dat <- den_proto %>%
  inner_join(data_frames, by = c("UBIGEO", "year", "month")) %>%
  mutate(date = ymd(paste(year, month, "01", sep = "-")))

vis_miss(dat %>%
           arrange(year, month))

write_csv(dat, "_data/LOR_dengue_clim.csv", na = "")

4.5 EDA

top8 <- dat %>%
  group_by(district) %>%
  summarise(cases = sum(den_cases)) %>%
  arrange(-cases) %>%
  slice(1:8) %>%
  select(district) %>%
  unlist()

capitals <- c("IQUITOS","YURIMAGUAS", "NAUTA", "RAMON CASTILLA",
              "REQUENA", "CONTAMANA", "BARRANCA", "PUTUMAYO")

plot <- dat %>%
  mutate_at(.vars = c(4, 11:20),
            .funs = ~scale_this(.)) %>%
  select(7, 21, 4, 11:20) %>%
  gather(var, val, 3:13) %>%
  filter(district %in% top8) %>%
  #filter(district %in% capitals) %>%
  filter(var %in% c("pdsi", "pr", "soil", "tmmx", "den_cases"))

plot_cases <- plot %>%
  filter(var != "den_cases") %>%
  inner_join(plot %>%
               filter(var == "den_cases") %>%
               select(-var), by = c("district", "date"))

plot %>%
  filter(var != "den_cases") %>%
  ggplot(aes(x = date, col = var, group = var)) +
  geom_line(aes(y = val), size =.3) +
  geom_line(dat = plot_cases, aes(y = val.y), col = "black", size =.3) +
  facet_wrap(factor(district, levels = top8)~var, 
             scales = "free_y", ncol = 4) +
  scale_x_date(date_breaks = "5 year", date_labels = "%Y")+
  theme_bw() +
  scale_color_npg() +
  theme(legend.position = "bottom") 

ggsave("./_out/time_series.png", width = 7, height = 10)