glmrob.nb version 0.4

This is glmrob.nb version 0.4. The following changes were made from version 0.3:

• Added the possibility for the design matrix X to be a single column full of ones, i.e. fitting a model only with an intercept.

glmrob.nb version 0.3

This is glmrob.nb version 0.3. The following changes were made from version 0.2:

• Added optional argument "param.ini" to bypass ML starting values and use user-supplied instead.

glmrob.nb version 0.2

This is glmrob.nb version 0.2. The following changes were made from version 0.1:

• Changed default values of tuning constants c.tukey.beta and c.tukey.sig to 5, as the previous ones were a little too likely to yield multiple solutions (depending on the design) which might have made the algorithm crash. Now, if the robust update of sigma given beta is unsuccessful (due to either multiple solutions or no solutions within the range defined by minsig and maxsig), the last value is kept and a warning is issued.
• All optional arguments are now documented in the html help file.
• Corrected some bugs related to how the weights on the design were passed on to some internal functions.
• Added optional arguments through a list in "options.wx" controlling some tuning aspects of the weights on the design (when not all set to one).

glmrob.nb version 0.1

This is glmrob.nb version 0.1. The following changes were made from the original version available at the Biometrics website on Wiley Online Library:

• Added explicit argument "offset" to the main function, with default value being a vector of ones. Argument referenced in help.glmrob.nb.html.
• Deleted handling of "..." in the main function until more tests assess that all relevant arguments are correctly passed to all glm() and lm().