In W-Mohammed/calibR: Calibration of Decision-Analytic Models

Progress report:

Tried calibrating:
A three state markov model:
- All methods managed to predict the targets.
- Adaptive MCMC MH failed to run due to singular values in the covariance matrix.
A three state markov model with a micro-Sim version:
- I had many issues running the optimisation algorithms. Most errors are due to a missing hessian matrix.
- The optimisation algorithm suggested implausible value for some of the parameter.
A hypothetical (dynamic) infections disease model:
- A few issues with missing hessian matrices.
- Implausible values for many of the parameters.
- Introduced transformations to log and logit scales to overcome the unboundedness feature of the used optimisation algorithms.
- Most optimisation algorithms proposed values that are significantly far from the target distribution.
- Apparently, the transformation of input parameters introduced bias into the Bayesian calibration process. This is likely the reason why most Bayesian results were off the charts.