Use of the "sample_obs" function (rbi package)

[Clara Champagne]

Hi all,

I am a new libbi user, using the rbi package and I had some unexpected results with the sample_obs function.
I attach an example with a deterministic SIS epidemic model re-fitted on simulated data.
It seems that the state variables are well estimated, but the sampled observations from the posterior (using sample_obs) are very far from the data (plot1).
I tried the function sample(bi, target="prediction", with="transform-obs-to-state") instead, and obtained completely different results (plot2).

Am I using these functions correctly? or do you have an idea why I observe such results ?

Thanks a lot for your help,
Clara

[Sebastian Funk]

Hi Clara,

I've pushed a fix (I think) to the rbi repository. Can you re-install from github and see if it improves things?

Seb.

[Clara Champagne]

Hi Seb,

This improves things a lot: it looks very good now!
Many thanks,

Clara