Skip to contents

Writing Vignettes

Source: vignettes/writing_vignettes.Rmd
writing_vignettes.Rmd

Introduction

This guidance ensures consistency across the vignettes in the admiral package in terms of content, structure and code execution. As a general rule, the vignette workflow defined in r-pkgs.org is followed.

Metadata

Each vignette in admiral should start with the following metadata.

---
title: "<Title>"
output:  rmarkdown::html_vignette:
vignette: >
  %\VignetteIndexEntry{<Title>}
  %\VignetteEngine{knitr::rmarkdown}
---

The <Title> of the vignette should be meaningful.

Markdown

Default Options

If any chunks are used within the vignette, the following options should be set after the metadata to ensure that the chunks are rendered consistently across all vignettes.

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

Format Sections

Table of Contents

Headings must be title case and start from Heading 1:

# Heading 1
This is the description of my first section.

## Heading 1.1
This is the description of my first sub-section within my first section.

## Heading 1.2
This is the description of my second sub-section within my first section.

The first section gives a brief introduction of the vignette. The last sub-section of the introduction should describe the packages required to run the admiral functions. The admiral package should be described first.

# Introduction

This is the introduction of my vignette.

## Required Packages
```{r, warning=FALSE, message=FALSE}
library(admiral)
# <all packages used in the vignette>
```

The warning=FALSE and message=FALSE options are there to prevent the usual messages:

Attaching package: ‘xxxx’The following objects are masked from ‘package:yyyyy’ fun1, fun2

Conventions

General Conventions

  • Any new vignette must be added to the _pkgdown.yml file in the relevant section.

  • Any variable name, dataset name, function, argument name must be quoted with backticks: e.g.

  The `date` parameter of the `derive_my_var()` function expects a date variable, e.g., `ADT`.

  • Functions must also end with ().

  • Variables and datasets name are expected to be uppercase.

  • All the codes must be described, executed and the output result displayed once the code is executed. Use:

    ```{r}
#<code>
```

  • Any output created must clearly show what the function has derived. It should at least show the variables/records used as input to the function and the derived variables/records. If a dataset must be displayed, it will be formatted using the dataset_vignette() function so that it is displayed consistently across the vignettes.E.g.

    • Description and execution of the code used to derive the variable/record
vs1 <- vs %>%
  derive_vars_dt(
    new_vars_prefix = "A",
    dtc = VSDTC,
    date_imputation = "first"
  )

Note: The call to get the formatted dataset would be:

  ```r
  dataset_vignette(
    vs1,
    display_vars = exprs(USUBJID, VSTESTCD, VISIT, VSDTC, ADT),
    filter = VSTESTCD == "WEIGHT"
  )
  ```

Displaying many big datasets on a vignette, may require a long time to load the page and may cause messages from the browser that the page is not responsive. In this case the number of displayed observations should be restricted either by restricting the source datasets at the beginning of the vignette or in the call of dataset_vignette() if only some calls are affected.

Using Footnotes

Footnotes can be useful to add context without adding clutter to the primary subject matter of the vignette being written.

To use footnotes, add a caret and an identifier inside brackets e.g. ([^1]). The identifiers can be numbers or words, but they can’t contain spaces or tabs.

The following markdown text will render as follows:

Here is an example [^1]

[^1]: Cool note about the example.

Here is an example 1

Conventions for ADaM Workflow

For vignettes describing an ADaM workflow,

  • the second section will summarize the programming workflow. The first sub-section within this workflow will always describe the data read to demonstrate the use of the admiral functions,

  • Each sub-section within the programming workflow should be tagged (e.g. [Step1] (#step)), so that the user can go on the relevant section from the programming workflow (in addition to the Table of contents). Don’t use a tag with a number but use a meaningful name (e.g. do not use (#link1), but use (#this_action))

  • the last section should link to a template script.

# Programming Workflow

* [Read in Data](#readdata)
* [Derive/Impute End and Start Analysis Date/time and Relative Day](#datetime)
* ...
* [Assign `ASEQ`](#aseq)

## Read in Data {#readdata}
## Derive/Impute End and Start Analysis Date/time and Relative Day {#datetime}
## ...
## Assign `ASEQ` {#aseq}

# Another Section

# Example Script
ADaM | Sample Code
---- | --------------
ADxx | [ad_adxx.R](https://github.com/pharmaverse/admiral/blob/main/inst/templates/ad_adxx.R){target="_blank"}

  • ADSL variables

    All ADSL variables which are required for any derivation are merged to the SDTM dataset before the first derivation. These ADSL variables have to be added to the by-variables for derivations which add observations. This ensures that the ADSL variables are populated for the new observations. A adsl_vars variable should be created at the beginning of the script and added to the by_vars parameter for derivations which add observations.

    ADSL variables which should be in the final dataset but are not required for any derivation are merged to the dataset after the last derivation.