To-dos

[LucileTerminassian]

• [x] Write code for the closed-form observational EIG

• [x] Write code for the EIG over causal parameters i.e. for the marginalization over non-causal parameters, using MCMC samples

• [x] Write code for the closed-form EIG over causal parameters

• [ ] Run BART and check that both EIG and EIG over causal parameters run smoothly with this model too

[Jakefawkes]

• [ ] Figure out why posterior from MCMC samples is not matching closed form posterior
• [x] Fix log likelihood func so it doesn't take an age to compute (shouldn't use inverse of large matrix but should instead sum individual likelihoods)