---
title: "Peak estimation"
author: "Tim Taylor"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Peak estimation}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.align = "center",
  fig.width = 7,
  fig.height = 5
)
```

```{r}
library(outbreaks)
library(incidence2)
library(i2extras) 
```

# Bootstrapping and finding peaks

We provide functions to return the peak of the incidence data (grouped or 
ungrouped), bootstrap from the incidence data, and estimate confidence intervals
around a peak.

## `bootstrap()`
```{r bootstrap}
dat <- fluH7N9_china_2013
x <- incidence(dat, date_index = "date_of_onset", groups = "gender")
bootstrap(x)
```

## `find_peak()`
```{r findpeak}
dat <- fluH7N9_china_2013
x <- incidence(dat, date_index = "date_of_onset", groups = "gender")

# peaks across each group
find_peak(x)

# peak without groupings
find_peak(regroup(x))
```

## `estimate_peak()`
Note that the bootstrapping approach used for estimating the peak time makes
the following assumptions:

- the total number of event is known (no uncertainty on total incidence);
- dates with no events (zero incidence) will never be in bootstrapped datasets; and
- the reporting is assumed to be constant over time, i.e. every case is
  equally likely to be reported.

```{r estimatepeak}
dat <- fluH7N9_china_2013
x <- incidence(dat, date_index = "date_of_onset", groups = "province")

# regrouping for overall peak (we suspend progress bar for markdown)
estimate_peak(regroup(x), progress = FALSE)

# across provinces
estimate_peak(x, progress = FALSE)
```