RECAP-Preterm wishlist

Support New functionality
1

We have a number of desires within RECAP which I think are mostly known about and/or covered by the wishlists posted by the LifeCycle and InterConnect projects. However, for the sake of completeness, I thought I would add what I currently know about our list (more potentially to be added by others).

Big desires are, in terms of analysis:

  • Multiple Imputation - I know this is currently being worked on*
  • Survival analysis - I believe that this is also being worked on*
  • Hierarchical modelling - I agree with @tombishop’s comment that this should be included within GLMM.

*When I say being worked on, I mean by the INESCTEC team in Porto; I hope they will contribute to the thread at some point.

From a data management perspective, I’d also agree that reshape and merge type commands would be useful.

Having said all of that, I think we might be a little behind the other projects in terms of establishing nodes etc (I may be wrong about that as I’m not really sure where LifeCycle is) and for the most part we are waiting to see (and use!) what new functionalities will be in the September release to really progress our analyses.

2

Hi Andrei,

Do you think you could prompt (Rui?) to comment on this, just to make sure we don’t duplicate efforts on survival analysis?

3

I Andrei and colleagues, as Andrei reffered at INESCTEC we are about to finish and document an R package for distributed computation of basic statistics, regression analysis and model checking (for DtataSHIELD) satisfying an initial RECAP project set of requirements. We have, as ongoing work, a distributed version of MICE’s imputation (first stage of MICE’s imputation is working) and also an implementation of “distributed survival analysis” (first step of the process). We have started (in collaboration with RECAP TNO colleagues) to look at the possibility of integrating traditional Machine Learning algorithms and Multi-relational Data Mining algorithms with DataSHIELD. For the latter case we are using the work o a colleague from the faculty of Sciences that allows to bind nicely Prolog and R :wink: The systems that we use for the relaltional DM are written in Prolog.

Also on the agenda (a suggesttion from RECAP partners), but not started ,is the distributed implementation of Structural Equation Modeling (SEM).

Best regards, Rui

1 Like
4

Thanks for the update, is the repository available on git hub (or somewhere else). I’m thinking that perhaps @demetris.avraam could take a look at these regarding integrating these into DataSHIELD.

5

We are about to finish testing and documentation (for the RECAP project users). I expect it will take a week more and then we put it in CRAN. I will make a post with details when we make it available.

Regards, Rui PS: There are a couple of ideas for improvements but we will incorporate them in an update version of the package to avoid further delays.

6

Hi Rui,

I was wondering if your package is available yet, as I would be very interested in having a look!

Thanks

Tom