BUGFIXES:

- examine function:
  Is that correct that also the check to see 'if there is no improvement in dose
  over the last cohort' should be added
  (Giuseppe's email) --> no, instead count number of times the same dose is 
  recommended contiguously and break after e.g. default 10000 times (can be option
  in examine function) to avoid infinite loop and issue a corresponding warning
  if this condition is met.

----------------------------------------------------------------
IMPROVEMENTS:

- add warning if the MinimalInformative prior or similar return a prior variance
  that is very low (to prevent undeliberately strong priors!)

- add a customized simulation summary function for the dual endpoint models,
  which also reports the frequency of simulated trials selecting the correct OBD

- include the submitted JSS manuscript as additional vignette in the package

- also optimize refDose for quantile->logistic functions

- make sure that it is clear in the documentation that the log transformation
  only applies to the slope (beta) and not to the intercept (alpha)
  
- the minimalInformative function should already produce a graph comparing the
  required and the resulting quantiles, instead of the user having to do it.

- record number of times the model is overruled in the simulations (Uli Beyer
request)

- mixture prior: add documentation in vignette

- allow different x values (doses) to be specified in simulate() to be passed to fit()

- remove WinBUGS since not really used anyway

----------------------------------------------------------------
NEW FEATURES:

- add slide production with ReporteRs package examples (to demo folder?)

- be able to run simulations on a cluster: use the BatchJobs package that is installed
  on the new Roche HPC - potentially as different fork of the package??

- production of Word output tables from simulation summaries

- minimal informative prior construction for other model(s), especially the probit model
  to allow construction for the dual endpoint models.

- other prior in the Kadane model -> how?

- historical data prior,
  pseudodata prior

