---
title: "Cost-Effectiveness Acceptability Curve Plots"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Cost-Effectiveness Acceptability Curve Plots}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r, include = FALSE}
knitr::opts_chunk$set(
collapse = TRUE,
comment = "#>",
fig.width = 6
)
```

```{r setup, results='hide', message=FALSE, warning=FALSE, echo=FALSE}
library(BCEA)
library(dplyr)
library(reshape2)
library(ggplot2)
library(purrr)
```

## Introduction

The intention of this vignette is to show how to plot different styles of cost-effectiveness acceptability curves using the BCEA package.

## Two interventions only

This is the simplest case, usually an alternative intervention ($i=1$) versus status-quo ($i=0$).

The plot show the probability that the alternative intervention is cost-effective for each willingness to pay, $k$,

$$
p(NB_1 \geq NB_0 | k) \mbox{  where  } NB_i = ke - c
$$

Using the set of $N$ posterior samples, this is approximated by

$$
\frac{1}{N} \sum_j^N \mathbb{I} (k \Delta e^j - \Delta c^j)
$$

#### R code

To calculate these in BCEA we use the `bcea()` function.

```{r}
data("Vaccine")

he <- bcea(eff, cost)
# str(he)

ceac.plot(he)
```

The plot defaults to base R plotting. Type of plot can be set explicitly using the `graph` argument.

```{r}
ceac.plot(he, graph = "base")
ceac.plot(he, graph = "ggplot2")
# ceac.plot(he, graph = "plotly")
```

Other plotting arguments can be specified such as title, line colours and theme.

```{r}
ceac.plot(he,
          graph = "ggplot2",
          title = "my title",
          line = list(color = "green"),
          theme = theme_dark())
```

## Multiple interventions

This situation is when there are more than two interventions to consider. Incremental values can be obtained either always against a fixed reference intervention, such as status-quo, or for all pair-wise comparisons.

### Against a fixed reference intervention

Without loss of generality, if we assume that we are interested in intervention $i=1$, then we wish to calculate

$$
p(NB_1 \geq NB_s | k) \;\; \exists \; s \in S
$$

Using the set of $N$ posterior samples, this is approximated by

$$
\frac{1}{N} \sum_j^N \mathbb{I} (k \Delta e_{1,s}^j - \Delta c_{1,s}^j)
$$

#### R code

This is the default plot for `ceac.plot()` so we simply follow the same steps as above with the new data set.

```{r}
data("Smoking")

he <- bcea(eff, cost, ref = 4)
# str(he)
```

```{r}
ceac.plot(he)

ceac.plot(he,
          graph = "base",
          title = "my title",
          line = list(color = "green"))
```

```{r}
ceac.plot(he,
          graph = "ggplot2",
          title = "my title",
          line = list(color = "green"))
```

Reposition legend.

```{r}
ceac.plot(he, pos = FALSE) # bottom right
ceac.plot(he, pos = c(0, 0))
ceac.plot(he, pos = c(0, 1))
ceac.plot(he, pos = c(1, 0))
ceac.plot(he, pos = c(1, 1))
```

```{r}
ceac.plot(he, graph = "ggplot2", pos = c(0, 0))
ceac.plot(he, graph = "ggplot2", pos = c(0, 1))
ceac.plot(he, graph = "ggplot2", pos = c(1, 0))
ceac.plot(he, graph = "ggplot2", pos = c(1, 1))
```

Define colour palette.

```{r}
mypalette <- RColorBrewer::brewer.pal(3, "Accent")

ceac.plot(he,
          graph = "base",
          title = "my title",
          line = list(color = mypalette),
          pos = FALSE)

ceac.plot(he,
          graph = "ggplot2",
          title = "my title",
          line = list(color = mypalette),
          pos = FALSE)
```

### Pair-wise comparisons

Again, without loss of generality, if we assume that we are interested in intervention $i=1$, then we wish to calculate

$$
p(NB_1 = \max\{NB_i : i \in S\} | k)
$$

This can be approximated by the following.

$$
\frac{1}{N} \sum_j^N \prod_{i \in S} \mathbb{I} (k \Delta e_{1,i}^j - \Delta c_{1,i}^j)
$$

#### R code

In BCEA we first we must determine all combinations of paired interventions using the `multi.ce()` function.

```{r}
he <- multi.ce(he)
```

We can use the same plotting calls as before i.e. `ceac.plot()` and BCEA will deal with the pairwise situation appropriately. Note that in this case the probabilities at a given willingness to pay sum to 1.

```{r}
ceac.plot(he, graph = "base")

ceac.plot(he,
          graph = "base",
          title = "my title",
          line = list(color = "green"),
          pos = FALSE)

mypalette <- RColorBrewer::brewer.pal(4, "Dark2")

ceac.plot(he,
          graph = "base",
          title = "my title",
          line = list(color = mypalette),
          pos = c(0,1))
```

```{r}
ceac.plot(he,
          graph = "ggplot2",
          title = "my title",
          line = list(color = mypalette),
          pos = c(0,1))
```

The line width can be changes with either a single value to change all lines to the same thickness or a value for each.

```{r}
ceac.plot(he,
          graph = "ggplot2",
          title = "my title",
          line = list(size = 2))
```

```{r}
ceac.plot(he,
          graph = "ggplot2",
          title = "my title",
          line = list(size = c(1,2,3)))
```


```{r echo=FALSE}
# create output docs
# rmarkdown::render(input = "vignettes/ceac.Rmd", output_format = "pdf_document", output_dir = "vignettes")
# rmarkdown::render(input = "vignettes/ceac.Rmd", output_format = "html_document", output_dir = "vignettes")
```