Changes since version 0.6.2
----------------------------
* Documentation updates: advice on AIC and citation to Kosmidis & Firth (2020, Biometrika).
* Added example on how to implement maximum penalized likelihood for logistic regression when the penalty is some power of the Jeffreys' prior (`?modifications`).

Changes since version 0.6.1
----------------------------
* Updated author and maintainer details.

Changes since version 0.5.8
----------------------------
* Removed dependence of print.glm on print.glm. Instead the generic method is called (thanks to Brian Ripley for contacting me on this).

Changes since version 0.5.7
----------------------------
* Implemented a fix for correctly resetting the "warn" option value to what the user has set prior to the execution of `brglm.fit` (thanks to Kurt Hornik for contacting me on this)
* Implemented a fix to avoid a warning related to the binding of n which is evaluated by family$initialize in `brglm.fit` (thanks to Kurt Hornik for contacting me on this)

Changes since version 0.5.6
----------------------------
* Minor changes in `example(modifications)` to avoid the use of `.Call` in the functions therein (thanks to Brian Ripley for contacting me on this).
* Updated contact information

Changes since version 0.5.5
----------------------------
* Added the `br.consts` argument to `brglm.control` which offers some handling of the starting values to the `brglm.fit` iteration.
* Updated contact and citation information.

Changes since version 0.5.4
----------------------------
* Improved constant adjustments for starting the fitting procedure in `brglm.fit`.
* Updated the references to papers in the help files.

Changes since version 0.5.3
----------------------------
* Minor corrections to the help files.
* The reported AIC, was incorrect, due to a minor bug. This is now corrected.
* Minor modifications to the separation.detection function
* Due to a bug, `brglm` was mishandling observations with zero weight. This is now corrected.
* Some improvements to profile and `confint` methods for `brglm` objects.
* Added citation file.


Changes since version 0.5.2
----------------------------
* Corrected a bug in `brglm` that could cause division by zero while scaling the design matrix.
* Due to a bug, `brglm.fit` was producing errors with model fits containing more that one aliased out parameters. This is now corrected (thanks to Aleks Jakulin for spotting the issue).
* Corrected a bug, that could affect the predict methods when aliased out parameters appeared (thanks to Aleks Jakulin for spotting the issue).

Changes since version 0.5-1:
----------------------------
* Correction of typos (sorry!) and various improvements of the help files.
