The logit transform:
- isn't well documented;
- is not in the user manual;
- has complications that aren't explained;
- has side-effects on printing the proposal covariance matrix which aren't documented;
- ditto 4. but for writing to a file.
Also the initial proposal covariance is mapped to a different domain to do the sampling, and we should describe this procedure as well.