https://liverpool-ac-uk.zoom.us/j/97622666757?pwd=RDg2OmTO0byxCy96qvEqgEAucfJbEJ.1
Reminder: Fill any DataSHIELD functionality developments in the shared spreadsheet: https://docs.google.com/spreadsheets/d/17OWWeLZH--RdCnp532oH6jpYib_Mg8O7/edit?usp=sharing&ouid=113438993511087013213&rtpof=true&sd=true
Attendees:
Apologies:
Agenda:
- Additional tags for the headers of DataSHIELD functions: disclosure filters, authors, contributors, maintainers.
- Auditing process independently to CRAN submission.
Notes:
REMEMBER TO USE THE NEW ZOOM LINK: https://liverpool-ac-uk.zoom.us/j/97622666757?pwd=RDg2OmTO0byxCy96qvEqgEAucfJbEJ.1
Attendees:
Stuart Wheater, Timm Intemann, Florian Schwarz, Nadja Lendle, Annika Swenne, Roy Gusinow, Andreas Mändle, Sofia Siampani, Tim Cadman, Manuel Huth, Ahmet Akkoç
Apologies:
Demetris Avraam
Agenda:
Notes:
- From December 2024 we will change the zoom link of our meetings to: https://liverpool-ac-uk.zoom.us/j/97622666757?pwd=RDg2OmTO0byxCy96qvEqgEAucfJbEJ.1
Protocol:
Topic 1:
- CRAN release of server-side package before client-side package because
CRAN Dependencies should not consist of any package to be installed from GitHub
- CRAN procedure very fast (automatic testing environment ~ 20min; E-Mail contact after 2 days)
- CRAN does not allow to have error, warnings or notes occurring during testing
Topic 2+3:
- dsBase(Client) release candidate 6.3.1 ready for release next week
- code coverage small in many community packages
Other Topics:
- Outreach: Community Map (Website + Raw Data for re-creation) of studies + people in DataSHIELD
- Outreach: Hex Stickers for DataSHIELD Packages
- Designer necessary (potentially Teresa Albers from NFDI4Health Consortium acc. to Timm I., interest also from Roy)
- important to come up with a feature shared across DS packages
- also indicates higher level of professionalism of DS Community
- Ahmet: there is an R (?) package for creation of such stickers
- Action: Florian will inform Outreach on topic
- DSI: Tim C. implemented new global option with Yannick so that DS server-side error is displayed nicely without having to call datashield.error() function
- dsSurvival Package (Timm I.): who is working on the package and what is the current status?
- Roche, Demetris (forked from Soumya into datashield repository)
- DSLite (Annika): Is it possible to run the servers in parallel instead of sequentially?
- unlear?
- Stuart will check this?
- Unassigned Action: we could need documentation on the difference in behaviour of DSLite in comparison to real-world setting
Attendees:
Manuel Huth, Florian Schwarz, Demetris Avraam, Annika Swenne,
Timm Intemann, Andreas Mändle, Roy Gusinow, Stuart Wheater
Apologies:
Agenda:
- Presentation: Andreas Mändle presented his work "A User-Friendly Interactive Dashboard for DataSHIELD".
- Updates from everyone for ongoing developments
- DataSHIELD versioning: What parts of the infrastructure need a version number (the entire system, the packages, an image, a profile)?
- Internal DataSHIELD functions: What is the best approach to develop/use internal functions?
- General updates
Notes:
Attendees:
Manuel Huth, Roy Gusinow, Demetris Avraam, Stuart Wheater
Apologies:
Agenda:
- General updates
Notes:
Attendees:
Ines Amine, Timm Intemann, Annika Swenne, Roy Gusinow, Stuart Wheater, Demetris Avraam, Andrei Morgan, Angela Pinot de Moira
Apologies:
Agenda:
- Presentation: Annika Swenne presented her work on “Federated Generalized Additive Models for Location, Scale and Shape in DataSHIELD”
- Stuart: dsDataShaper, pipelines for data shaping
- Andrei: Use wiki for agenda and meeting minutes
- General updates
Notes:
Attendees:
Manuel Huth, Timm Intemann, Andreas Mändle, Florian Schwarz, Annika Swenne, Roy Gusinow, Angela Pinot de Moira, Andrei Morgan, Stuart Wheater, Paula Irving
Apologies:
Demetris Avraam
Agenda:
- Presentation: Manuel Huth presented his work on “Challenges with developing non-linear mixed-effects models in DataSHIELD”
- Update on cox development for ProPass (did not happen)
- General updates
Notes:
Attendees:
Demetris Avraam, Becca Wilson, Florian Schwarz, Stuart Wheater
Apologies:
Annika Swenne, Andreas Mändle, Timm Intemann, Andrei Morgan, Manuel Huth
Agenda:
- Continue the discussion for the audit process
- Public roadmap - what mechanism: wiki, github, embedded spreadsheet?
- Cox Regression
- Conference speakers
- General updates
Notes:
- Manuel Huth and Camille Bachot to present in the next two meetings
- Does Stuarts testing incl version R
- release note re version R
Attendees:
Demetris Avraam, Tom Bishop, Manuel Huth, Florian Schwarz, Timm Intemann, Becca Wilson, Paula Irving, Angela Pinot de Moira, Roy Gusinow, Andreas Mändle, Sofia Siampani, Miron, Aikaterini Lymperidou, Stuart Wheater
Apologies:
Annika Swenne, Andrei Morgan
Agenda:
- Presentation: Florian Schwarz presented “Creating and Audit Process for DataSHIELD Packages” (20mins + 10 mins discussion)
- Takeaways from the conference - Audit discussion
- Becca: draft of UK guidelines for output checking
- Becca: differential privacy package
- Demetris: DataSHIELD wiki and Tech Group meetings
- General updates
- DataSHIELD Community calendar
Notes:
- Manuel Huth to present in the first meeting of 2024 and Camille Bachot to present dsPrivacy in the second meeting
- To continue the discussion for the audit process in the next theme meeting
Attendees:
Demetris Avraam, Stuart Wheater, Manuel Huth, Tim Cadman, Roy Gusinow, Florian Schwarz, Annika Swenne, Andreas Maendle, Andrei Morgan
Apologies:
Becca Wilson, Timm Intemann
Agenda:
- Presentation: Tim Cadman presented his work on “Developing tidyverse functions in DataSHIELD” (20mins + 10 mins discussion)
- Developing 1-stage Cox ph functions in DataSHIELD
- General updates
Notes:
- Tim will look for a group of ds.tidyverse -> WikiEntry
- Demetris, Roy & Manu will work on ds.coxReg (name debatable)
- Florian: We should summarize the takeaways from the conference -> next meeting’s Agenda -> collect ideas
- Andrei: Post minutes of the meetings to the wiki
Attendees:
Demetris Avraam, Manuel Huth, Annika Swenne, Stephan Ringshandl
Apologies:
Carolina, Roy
Agenda:
- General updates
- Parallelization of ds.base
- Demetris + Manuel: Discussion on what to send to Andrei for the conference
Notes:
- work on parallelization of ds.base (dplyr, no for loops, compute linear model directly, checks can be turned off, prints can be turned off)
Attendees:
Stuart Wheater, Timm Intemann, Stephan Ringshandl, Florian Schwarz
Apologies:
Demetris Avraam, Annika Swenne, Andrei Morgan
Agenda:
- General updates
- Homomorphic encryption capabilities via HomomorpheR -> Shared private key among data owners
Notes:
- Stuart: started having a weekly Drop-In-Session / Meeting for the Operational Theme (Wednesday)
- Florian: work related to the cluster analysis function (specifically related to dsBase 6.3 version; changes in the permission mode) and microbiome functions
- Timm (on behalf of Annika): first GAMLSS model implementation in DataSHIELD
Attendees:
Demetris Avraam, Manuel Huth, Annika Swenne, Andreas Mändle, Stephan Ringshandl, Beeca Wilson, Carolina Alvarez Garavito, Stuart Wheater
Apologies:
Andrei Morgan, Roy Gusinow, Florian Schwarz
Agenda:
- Discussion: Plan for dsBase 7.0 release
- General updates
Notes:
- skip meeting in August
- dsBase release: Improve performance of dsBase functions (maybe set option to exclude checks; make more use of dplyr; directly add columns to data frames or replace columns directly, make use of sapply/sapply instead of for loops)
- Include checks in scatter plots that were described by Stephan
- Stephan: Multiple servers are executed/called after each other -> can this be done in parallel? Answer: Can be done in Client library in assign functions)
Attendees:
Demetris Avraam, Becca Wilson, Carolina, Stephan Ringshandl, Manuel Huth, Timm Intemann, Annika Swenne, Florian Schwarz, Stuart Wheater, Hank
Apologies:
Agenda:
- Presentation: Stephan Ringshandl presented his work on “Brute Seed disclosure method in DataSHIELD” (20mins + 10 mins discussion)
- Discussion: Roadmap for dsBase 6.4 (or 7.0)
- (if possible: discussion HP Swarm Learning)
- General updates
Notes:
Attendees:
Demetris, Christian, Andrei, Stuart, Becca, Manuel, Roy, Carolina, Hank
Apologies:
Agenda:
- Presentation: Manuel Huth presented his work on “Privacy-preserving Difference-in-Differences with multiple time periods” (20mins + 10 mins discussion)
- Discussion: Separate functions (more packages + papers? -> marketing) or include “all” functions in dsBase?
- Discussion: Roadmap for dsBase 6.4 (or 7.0)
- General updates
Notes:
Attendees:
Demetris, Roy, Stuart, Florian, Hank, Manuel
Apologies:
Andrei, Becca, Juan, Tom, Tim
Agenda:
- Presentation: Demetris Avraam presented his development of the ds.mice function for multiple imputations in DataSHIELD.
- General updates
- Miscellaneous (Continuous Integration, Being up-to-date with standard R packages [dplyr -> issue in ds.make with matrices, tidyverse, etc.], summary and plot functions for ds.glm and other model classes [standards] -> summary(ds.glm.object yields just a list summary?); ds.glm prints vcv automatically [why?]; standard naming convention?)
Notes:
- continuation of ds.mice package functionalities (Demetris)
- global version for imputation would be desirable (time consuming)
- updates: Manu -> ds.did package; Florian -> conferences marketing with DataSHIELD; Stuart -> ds.base release, docker images for analysis with multiple packages; Hank -> new version of ds.MTL out with automatic capture of covariance during model Training, request: remove automatic printing of functions; Roy ->
- presentations: ~ 20 mins + ~ 10 discussion
- Manu: should we stick to CRAN as much as possible when creating model objects? (considering disclosure, of course)
- Next presentation by Manu or Roy
- Split up dsBase package in multiple packages (dedicated meeting for this) -> good from a technical and a marketing point of view
Attendees:
Demetris Avraam, Manuel Huth, Dick Postma, Augusto Anguita, Christian Hilbrands, Florian Schwarz, Becca Wilson, Stuart Wheater
Apologies:
Tom Bishop, Tim Cadman, Roy Gusinow
Agenda:
- Introductions
- How to organize the Theme meetings
- Statistical developments plan
- Working groups
Notes:
- One meeting per month with a short presentation and discussions, and then discuss any issues from developers
- We can suggest a general Drop-in session for the whole Core Tech Group
- We can have drop-in sessions for users (we can discuss this also with Juan)
- To have an image on all available type of analysis, from all tested community packages
- Spreadsheet to be shared between members of the theme. At later stage we can create a live road map shared publicly with everyone, where the users can see what we plan and when in future releases
- Group on continuous integration and automated package testing