- optimize automatic proposal distribution generation
- enable fixing the guessing rate to an arbitrary value (? which?)        <- This is actually possible by putting a strong prior on the guessing rate...
- what is going on with type checking in psigniplot.GoodnessOfFit?
- add a dictionary to the data objects to make them know what they are (stimulus intensity, ...)
- ROC curves for 1AFC
- allow for the gamma=lambda prior

- Blob size reduction for PF
- Deviance Bootstrap only one sided test
- Error bars on PF. Simply streched bars? (Not sure what is meant)
- 68 Confidence Interval
- In 3d/4d, we could also integrate the posterior without monte carlo methods. Advantage: We get a somehow analytical formula for the posterior. Disadvantage: Might still take quite long and complicates things like CI estimation. In addition, it would be difficult to estimate the errors originating from the fact that we approximate |R by a compact interval.
- Plot MCMC diagnostics  for all parameters (Not sure what is meant)
- Plot the blue line on top of all the white ones, in the diagnostic plot.
- Alternative view: Shaded region of the posterior instead of 20 sample PFs
- In diagnostic plots, evtl. ad a message on top (what does this mean)
- Replace "model correction" with psi(x) or psi(stimulus intensity) (for people with less elaborated statistics background)
- Influence for Slope and Influence for Threshold seperately instead of using an
  aggregation in the Plot for the Influential Observers.
