---
title: "-4- Compute Confidence Interval"
output:
  rmarkdown::html_vignette:
    toc: true
    toc_depth: 4
description: >
  This vignette shows how to use GroupSeq to compute confidence intervals
  at the final stage of a group sequential analysis.
vignette: >
  %\VignetteIndexEntry{-4- Compute Confidence Interval}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r knitr-setup, include = FALSE}
library(knitr)
knitr::opts_chunk$set(
  comment = "#",
  prompt = F,
  tidy = FALSE,
  cache = FALSE,
  collapse = T
)

old <- options(width = 100L, digits = 10)
```

Let's assume you have planned a two-sided 3-stage Pocock design with 80% power.

```{r, out.width = "35%", echo = FALSE, fig.alt = "two-sided-three-stage-pocock-setup"}
include_graphics("figures/task4-two-sided-3-stage-Pocock-setup.png")
```

The resulting design requires a drift of about 3.

```{r, out.width = c("55%", "38%"), echo = FALSE,fig.show="hold", fig.alt = "two-sided-three-stage-pocock-result"}
include_graphics("figures/task4-two-sided-3-stage-Pocock.png")
include_graphics("figures/task4-two-sided-3-stage-Pocock-graph.png")
```

During the study you perform the interim analysis as planned when about 67% of
all samples were collected and obtain a standardized effect size of 2.5, exceeding
the critical bound (2.2894) so that you can abort the study and reject H0.

To compute the confidence interval at this point, open option -4- of GroupSeq and
enter the values as shown below.

```{r, out.width = "45%", echo = FALSE, fig.alt = "task-4-CI-setup"}
include_graphics("figures/task4-CI-setup.png")
```

Hitting `CALCULATE` yields the following.

```{r, out.width = "50%", echo = FALSE, fig.alt = "task-4-CI-result"}
include_graphics("figures/task4-CI-result.png")
```

So for this study, the resulting confidence interval of the standardized
effect is (0.30, 5.38).

Depending on the underlying data distribution and applied
test statistic this will be "back-calculated" to obtain the confidence interval
for the value of interest (e.g. mean difference for some normally distributed
data or difference in proportions of some binomially distributed data).


```{r, include = FALSE}
options(old)
```