First round of inference using PyMC3 with Nuts

[fccoelho]

First attempts at inference with simulated data with lognormal distribution for SIR with 2 and 1 equation.

[maxbiostat]

Here's something I don't get from this notebook: are we placing a prior on both R_0 and (gamma, beta)?

Or is the prior on R_0 only used for the single equation model and the joint prior on the rates used for the "two-equation model"?

[fccoelho]

The prior on R0, is to constrain the system to an epidemic state, i.e. it forces beta to be greater than gamma.

[maxbiostat]

Makes sense, but would be nice to have two runs for the single ODE model:

• the induced (log-normal) prior by the (log-normal) priors on the rates
• the truncated log-normal prior you're already using.

We are then interested in differences in

• Inferential summaries: mean, median, BCIs
• Divergences
• Run time