How GGIR became the software of choice for analysing activity monitors

26 Sep 2019 - 4 min

eScience Engineer Vincent van Hees on GGIR, one of the most widely used methods to analyse modern activity monitors data

Since 2008, wrist worn activity monitors that store high-resolution output have been increasingly employed to study human behaviour and movement. One of the most widely used methods to analyse modern activity monitors data is GGIR, an open source R software package developed by Dr Vincent van Hees, Senior Research Engineer at the eScience Center and independent consultant.

In the following interview, Van Hees explains the benefits of GGIR and how this software package became so successful.

What is GGIR?

The research softwareI, with the help of others, developed for health research to process and analyse data collected with wearable movement sensors (accelerometers).

How did this software package come about?

During my PhD in Cambridge, I developed a range of algorithms for processing wearable movement sensor data and I started to apply these algorithms todata sets collected under real lifeconditions.Initially, I did this with my own local R scripts, but this soon evolved into anopen source R package applicable to the wide diversity of study designsinvolving a variety of sensor brands.

GGIR loads the data form files, corrects for sensor calibration error, extracts signal features, provides estimates of daily sleep and activity patterns and generates detailed quantitative summaries.In addition, it provides a couple of simple visualisations to facilitate data-quality checks.

GGIR continues to be widely used by researchers across the world. How would you explain its popularity?

I think the popularity comes from a number of factors:

  • The software is broadly applicable within the research field.
  • Open source software is “free” (not development time of course)
  • The online resources to guide new users seem sufficient.
  • The software is writing in the R language, which makes it fit well with the typical user.
  • The software is used in various prestigious data sets and publications that act like a billboard for the software and add to the credibility of the software.
  • The software was the first of its kind in the field, which gives it a competitive advantage.
  • And the popularity fuels itself because the research community favors methodological consistency.

Would you say working at the eScience Center has benefited the development and maintenance of GGIR? Why?

Yes, at a personal level working at the eScience Center has helped me to strengthen my software development skills and broaden my understanding of data analysis techniques.Moreover, it has made me aware of the challenges across scientific domains in software development and sustainability, and about existing solutions. The software quality of GGIR has benefitted immensely from this. For example, by addingversion control, unit tests, and continuous integration it is much easier for others to make a contribution to the code and much easier for me to secure consistent functionality.Also, the eScience Center has acted as host organisation for three externally funded projects to work on new algorithms for GGIR. These projects have boosted the development of GGIR.

GGIR logo

This popularity eventually led you to switch careers and become an Independent software consultant. What made you decide to do this?

In 2018 I started doing independent consultancy for half a day a week, which I later that year expanded to one full day per week. This exploratory phase provided me with the confidence to work as an independent consultant.I have been building expertise around movement sensor data analysis since 2003 and being able to keep usingall this experience feels good.

What are your plans with GGIR for the coming years? Any other projects you are working on?

Thousands of researchers around the world are using wearable movement sensorsin their research and need open source software solutions for a wide range of data challenges. Sofor the moment, there is enough workto do on software and algorithm developmentand on training software users.Additionally, I am collaborating with researchers at the Southern Danish Universityin Odenseon building a platform for international pooling and analysisof wearable movement sensor and GPS data.

Also, I see an increasing demand for translating open source software solutions supported by academiato applicationswithin the health care industry. For that reason, I have partnered up with Shimmer and Nextbridge health (US) to develop services around open source software in the health care industry.

Watch the GGIR tutorial