Building composite indicators with {composer}

Author

William Becker

Published

January 24, 2024

1 Welcome

Welcome to the documentation for the {composer} app. The {composer} app provides an easy interface for building composite indicators using any data set. Users can visualise and explore results in depth, and download figures and reports.

{composer} is a Shiny app which is wrapped in an R package. At the moment, in order to run the app you will have to have R installed, then first install the package:

# install package if not already installed
remotes::install_github("finddx/composer")

To run the app you then have to run:

# load package
library(composer)

# run app
run_gui()

As a Shiny app, the app will open in your web browser and acts as an interactive GUI.

The rest of this page gives some general information about the app and composite indicators. If you want to get started quickly, go straight to Chapter 2. Detailed documentation on each tab of the app starts at Chapter 4.

1.1 Composite indicators

Indicators are used in many contexts to measure complex multidimensional concepts, typically with the aim of prioritising resources and interventions, and also to track progress. In the international/policy context, indicators are often used to compare countries and/or sub-national regions. Examples include the Human Development Index, the Global Innovation Index, and many others.

Quite often, the concept to be measured cannot be sufficiently captured with one indicator, and a group of indicators is needed. As the number of indicators gets larger, it becomes increasingly difficult to compare and prioritise.

Composite indicators are mathematical aggregations of a set of indicators into a single measure. Indicators are organised into conceptual groups which aim to follow a map of the concept to be measured. Aggregating the indicators into a single composite indicator allows quick and easy comparisons, clear communication with stakeholders, and acts as a natural entry point to the data set underneath.

Importantly, in building a composite indicator, we do not wish to substitute the underlying data, but rather to complement it with an overview measure. Composite indicators involve a number of subjective decisions in their construction, and cannot fully capture all information in the indicator set underneath. However, used carefully, they are a valuable addition and entry point to a complex data set.

1.2 Features

The {composer} app includes the following features:

  • Any number of indicators and units and aggregation levels
  • Unit screening by data availability
  • Missing data imputation
  • Outlier treatment
  • Normalisation using various methods
  • Weighted aggregation
  • Interactive maps (if the units are countries)
  • Detailed analysis of indicators using bubble charts, bar charts
  • Statistical analysis using visualisation of distributions, correlation plots
  • Downloadable unit profiles
  • Interactive reweighting
  • Sensitivity analysis on assumptions, checking the effects of removing indicators and indicator groups

1.3 Methodology

The composite indicator methodology used in this app follows the internationally-recognised OECD/JRC Handbook of Composite Indicators, and the methodology used by the European Commission. If you wish to learn more about building and analysing composite indicators, see for example the recent training course given by the European Commission, including videos and slides.

Most of the data processing in the app is done using the COINr package, which is an R package for building and analysing composite indicators. R users may wish to additionally work with COINr directly to access functionalities not included in the app.

1.4 Terminology

Throughout this book we will use the following terms:

  • Units are the set of “things” that we want to compare and possibly rank. Often units are countries, but they could also be regions, cities or organisations, which is why we use “units” here.
  • Indicators are measured variables which we use to compare units.
  • A composite indicator is a mathematical aggregation of a set of indicators into one composite measure, which aims to summarise the indicator set as well as possible.

1.5 Development and contribution

{composer} was developed over 2023-2024 and (at the time of writing in Feb 2024) is in an early release phase. If you find a bug in the app, or have a feature request, please open an issue on GitHub. Collaboration and contribution are welcome: if you wish to contribute please fork the GitHub repo, make any changes and make a pull request. Feel free to discuss on the issues page, or contact us.

1.6 Acknowledgements

{composer} was originally built as an internal-use app for the Foundation for Innovative New Diagnostics (FIND), who then kindly granted permission for an open-source release. The app is effectively a front end for the COINr package, and also benefits from many other open-source R packages.