Abstract submission to propose discussion sessions, showcase usecases or propose/demonstrate developments is also open: www.datashield.ac.uk/abstract2019. Closes 31st July 2019.
Please note this meeting is co-located with the EUCAN-Connect AGM which will run 9-10th September.
There’s currently only a brief description on the abstract page saying that abstracts are being accepted. I’ve probably completely missed it somewhere else, but is there a bit more information available on what format is required and how to submit them? Thanks.
@becca.wilson thanks - I hadn’t realised the webform was there: it turns out it was getting blocked by teh Privacy Badger addon which I have installed.
There’s currently only a brief description on the abstract page
saying that abstracts are being accepted. I’ve probably completely
missed it somewhere else, but is there a bit more information
available on what format is required and how to submit them? Thanks.
I Andrei,
Last year we register with this information:
Name Rui Camacho
Address FEUP Rua Dr Roberto Frias, s/n
Porto, 4200-465
Portugal
Email rcamacho@fe.up.pt
Submission Type Standard talk (20 mins + 10 mins questions)
Abstract Title Distributed Statistical Analysis Using DataSHIELD in
Epidemiological Studies
Authors and affiliations
Paula Raissa , INESC TEC
Gonçalo Gonçalves, INESC TEC
Carla Moreira, ISPUP
Vitor Santos, INESC TEC
José Pedro Ornelas, INESC TEC
João Correia Lopes, INESC TEC & FEUP
Artur Rocha, INESC TEC
Rui Camacho, INESC TEC & FEUP
Abstract (300 words max)
DataSHIELD is a set of R packages, that enable data analysis in a
distributed setting without the need to disclose or transfer
individual-level data. DataSHIELD already provides some packages to make
statistical analysis like generalized linear models, tests and basic
statistics calculations. For more extensive and in depth statistical
analysis we need to extend DataSHIELD repertoire of data analysis
functions. We have developed a new R package that includes more
regression models, basic statistics and Goodness-of-fit tests. Our
approach to build the R package for distributed analysis, required the
rewriting of the statistical formulae by formula decomposition. The
reformulation of the formulae allows for the distributed computation of
aggregated values (summations basically) on ech data nodes. Final values
are then computed at a central node using aggregated data only. We also
aimed at minimizing the amount of calls to each data node. The proposed
strategy yields the same results as polling data analysis. For
regression models, our package includes functions like linear, logistic,
poisson, polynomial and hierarchical regression. For basic statistics
our main contribution is the conditional percentiles functions. For
model checking there are functions like Anova, t-Test, Kruskal Wallis
and others. We are currently developing important functions on Survival
analysis, such as, Kaplan Meier, Nelson Aalen and Cox Hazard.
The R package was tested using a small sample of Generation XXI cohort
data set. The package was also tested with a set of artificial data sets
specially designed to test efficiency and robustness. For demonstration,
we will show a pilot study using Generation XXI data set and conditional
percentile functions.
The work was done under the RECAP preterm project that is funded from
the European Union’s Horizon 2020 research and innovation programme
under grant agreement No 733280
I will ask Paula the details, but I think that it was a single PDF
file.
Regards,
Rui
We want to have a session where each cohort study (LifeCycle, Interconnect and Recap) can give a 10 mins introduction/overview of their consortia (number of sites etc), what they are researching and also how/when they will/are using DataSHIELD. I think this will be helpful for delegates within and external to EUCAN-Connect. Is it possible for @tombishop, @sidohaakma@asm to liaise with their respective consortia and let me know who will be able to deliver this so I can work it into the agenda.
Hi @becca.wilson, sure RECAP Preterm can present something short, that would be good. I think we’re still figuring out who is going to attend so let’s stay in touch about this. Cheers, Andrei