The Statistical Development Theme serves as the central hub for coordinating the ongoing development of DataSHIELD functions and packages, playing a pivotal role in the evolution of privacy-preserving data analysis using DataSHIELD.
Theme Leads: Demetris Avraam and Manuel Huth
Our overarching vision is to continuously expand and maintain the software package libraries of DataSHIELD. Our primary focus is on the methodological interface of programming and Federated Learning. On the development front, we are dedicated to crafting novel packages for federated statistical analysis while enhancing existing DataSHIELD software packages. We promote an open-science culture and actively guide users, enabling them to contribute to DataSHIELD by creating their own software packages.
The Statistical Development Theme convenes on the first Thursday of each month. If you wish to participate, kindly join us at 10:30 AM (CET) via the Zoom link. No registration is required. You can find the minutes of our past meetings, as well as the agenda for the next meeting, here.
During these gatherings, one of our developers or users presents their latest work results and ongoing progress. We place tremendous value on feedback from both our developers and users, as it is the key to enhancing our suite of software tools. If you have work you'd like to showcase and receive valuable feedback on, we encourage you to get in touch with us! We are happy to give you a presentation slot soon.
We are always open to collaboration on our active projects. If you're developing a package and wish to have it featured below, please contact us with a brief description, and we'll be happy to showcase your project.
Current projects include:
In this project, we plan to develop DataSHIELD versions of a limited number of dplyr functions to aid data manipulation. We will also explore the feasibility of implementing Tibbles and the pipe operator in DataSHIELD. Anyone interested in contributing should contact Tim Cadman (tica@sund.ku.dk)
For this initiative, our goal is to integrate the renowned Cox survival analysis model with DataSHIELD. Demetris Avraam (demetris.avraam@sund.ku.dk) is leading the project and welcomes collaborators. Potential contributors can assist in areas such as code development, methodological discussions, and model testing.
In this initiative, we are focused on introducing federated non-linear mixed-effect models into DataSHIELD. Manuel Huth (manuel.huth@helmholtz-muenchen.de) is leading the project and is keen on welcoming potential collaborators. Interested parties can contribute in various capacities, including code development (mainly translating Julia code to R and DataSHIELD and adding new distributions) and model validation.
We and the Operational Management Theme have now restarted a weekly DataSHIELD 'drop-in' group. This group is open for everyone to attend and will be facilitated by Demetris Avraam, Stuart Wheater and Tim Cadman. The aim is to provide a setting where people can bring questions/issues with DataSHIELD code or analysis and we will try to help! We can also offer support with operational issues/queries.
You can join the meetings every Wednesday at 10 am (CET) without registration via this zoom link.
Should you have any inquiries or suggestions, please don't hesitate to reach out to either Manuel Huth (manuel.huth@helmholtz-muenchen.de) or Demetris Avraam (demetris.avraam@sund.ku.dk). Your input is highly appreciated and vital to our collective mission of privacy-preserving data analysis.