Model checking for simplified Naomi

Abstract

Background We have run the simplified Naomi model using a range of inference methods.

Task We check the appropriateness of the model for the data.

tmb <- readRDS("depends/tmb.rds")

outputs <- tmb$outputs

indicators <- naomi::add_output_labels(outputs) %>%
  left_join(outputs$meta_area %>% select(area_level, area_id, center_x, center_y)) %>%
  sf::st_as_sf()

## Joining, by = c("area_level", "area_id")

indicators %>%
  filter(
    age_group == "Y015_049",
    indicator == "prevalence",
    area_level == 4
  ) %>%
  ggplot(aes(fill = mode)) +
  geom_sf() +
  viridis::scale_fill_viridis(labels = scales::percent_format()) +
  naomi::th_map() +
  facet_wrap(~sex)

indicators %>%
  filter(
    age_group == "Y015_049",
    indicator == "prevalence",
    area_level == 4
  ) %>%
  names()

##  [1] "area_level"       "area_level_label" "area_id"          "area_name"        "sex"              "age_group"        "age_group_label"  "calendar_quarter" "quarter_label"    "indicator"        "indicator_label"  "mean"            
## [13] "se"               "median"           "mode"             "lower"            "upper"            "center_x"         "center_y"         "geometry"

Next steps here are:

• Add column to indicators with the “raw” value of the indicator
• Add column to indicators with the quantile of the raw value within samples from the posterior
• Start outputting tmb$outputs for the other inference methods
• Use existing or own functions to produce coverage histograms and ECDF difference plots