metric-details.Rmd

---
title: "Details on the metrics implemented in `scoringutils`"
author: "Nikos Bosse"
date: "`r Sys.Date()`"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Details on the metrics implemented in `scoringutils`}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)
library(scoringutils)
library(kableExtra)
library(magrittr)
library(knitr)
library(data.table)
```

## Applicability of different metrics for different types of forecasts and formats

This table gives an overview for when which metric can be applied and gives a very brief description. Note that this table on shows the metrics as implemented in `scoringutils`. For example, only scoring of sample-based discrete and continuous distributions is implemented in `scoringutils`, but closed-form solutions often exist (e.g. in the `scoringRules` package). 

```{r, echo = FALSE, results = "asis"}
data <- copy(metrics)
setnames(data, old = c("Discrete", "Continuous", "Binary", "Quantile"),
         new = c("D", "C", "B", "Q"))
data[, c("Name", "Functions") := NULL]

replace <- function(x) {
  x <- gsub("+", "y", x, fixed = TRUE)
  x <- gsub("-", "n", x, fixed = TRUE)
  return(x)
}

data$D <- replace(data$D)
data$C <- replace(data$C)
data$B <- replace(data$B)
data$Q <- replace(data$Q)

data[, 1:6] %>%
  kbl(format = "html",
      escape = FALSE,
      align = "lccccl",
      linesep = "\\addlinespace") %>%
  column_spec(1, width = "3.2cm") %>%
  column_spec(2, width = "1.5cm") %>%
  column_spec(3, width = "1.5cm") %>%
  column_spec(4, width = "1.3cm") %>%
  column_spec(5, width = "1.5cm") %>%
  column_spec(6, width = "6.0cm") %>%
  add_header_above(c(" " = 1, "Sample-based" = 2, "  " = 3)) %>%
  row_spec(seq(1, nrow(data), 2), background = "Gainsboro") %>%
  kable_styling()
```


## Detailed explanation of the metrics implemented in `scoringutils`

```{r, echo = FALSE, results = "asis"}
data <- readRDS(
  system.file(
    "metrics-overview", "metrics-detailed.rds", 
    package = "scoringutils"
  )
)

data[, 1:2] %>%
  kbl(format = "html",
      escape = TRUE) %>%
  column_spec(1, width = "3.5cm") %>%
  row_spec(seq(1, nrow(data), 2), background = "Gainsboro") %>%
  column_spec(2, width = "15.5cm") %>%
  kable_styling()
```