A week in Barcelona (2)

ISPOR conference is nearly over — still a couple of interesting sessions to go. Yesterday we did our session on missing data (not sure why my name doesn’t appear there — but believe me, I was there and gave my talk! I think the slides will be available soon). And I did go to a couple more interesting sessions, including one on extrapolation of survival outcomes.

Today I am going to be at another session on what is probably become my main windmill crusade — that’s about moving practitioners towards the use of proper and appropriate statistical software for modelling.

This is of course related to the workshop we did last summer at UCL and I do think we are gaining some traction — Jeroen, Devin, Howard and I have also written an editorial, which has received very positive reviews (and )we should be able to address all the comments and hopefully have it published in Value in Health soon).

All in all, I think I’ve enjoyed the conference and there’s lots of interesting ideas to bring home — one that I think I should really find time to work on very soon is to implement in BCEA the reporting facilities available in the webapp. Won’t be too complex to add a function so that would allow the user to create a full report with all (or a selected subset of) the analyses — something like

makeReport(analyses=c("decision_analysis","psa","evppi"),type="doc",name="report.doc")

with also some “standardise” code embedded in the report. Much of it is already available in the webapp so it’d be mostly finding time to translate it to the format of the package.

Another interesting idea, which I discussed with Nikos is to check out the Stan Variational Inference algorithm — particularly for the survHE more computationally expensive models (e.g. those based on splines). Again shouldn’t be too difficult to embed this additional option to the main fit.models(...) function.