25 Nov 2023: dglm 1.8.6

- Remove examples sections from anova.dglm.Rd, dglm.object.Rd and
  summary.dglm.Rd because they are identical to the example in
  dglm.Rd.
- Update URLs for references to use DOIs and to use gksmyth.github.io
  instead of www.statsci.org.
- Other minor edits to summary.dglm.Rd.
- Convert itemized list of components to table in dglm.object.Rd.
- Complete rewrite of Tests.R. Set the seed for reproducibility and
  save the output file. Rename Tests.Rout to Tests.Rout.save.
- Add https://gksmyth.github.io/s/dglm.html URL for legacy code to
  this ChangeLog.

19 Dec 2022: dglm 1.8.5

- Capitalize help file titles.
- Convert aliases for dglm.object.Rd to be that for docType{class}.
- Use American spellings.
- Import instead of depend on statmod package in DESCRIPTION.
- Add export() to NAMESPACE and remove exportPattern().
- Add importFrom() to NAMESPACE to explicitly import functions from
  the stats and statmod packages. Remove import(statmod) from
  NAMESPACE and remove `stats::` from code.
- Add Test.Rout file to test directory.
- Update and complete ChangeLog file.
- Update references to Verbyla & Smyth (1999).

05 Nov 2020: dglm 1.8.4

- Package maintenance taken over by Gordon Smyth.
- Remove uses of sys.parent(), which are not allowed by CRAN.

24 Aug 2016: dglm 1.8.3

- Delete unused arguments in anova.dglm().

04 Nov 2015: dglm 1.8.2

- Package maintenance taken over by Robert Corty
  <robert_corty@med.unc.edu>.
- Added model frames to the return object as [["model"]]. This is consistent with how 
  glm() works and fixes a bug where predict(dglm.object, se = TRUE) produced an error 
  when called from any environment other than global.

07 Nov 2014: dglm 1.8.1

- Fixed an error where mu<0 was flagged as an error, even for the Gaussian case.
  The code now checks that the dispersion is greater than zero, the original intent.
  This was the cause of producing some errors in output seemingly unrelated.
  Thanks for Lars Ronnegaard and Gordon Smyth for identifying this error.
   
30 October 2012: dglm 1.6.2
	
- Minor edits to code and help files.
- Fix minor error in dglm.control.Rd.
- Dispersion model was incorrectly fitted (even in the examples)
  compared to the given specification; introduced a fix that is
  not entirely satisfactory, but works.

06 May 2006: dglm 1.5
 
- Corrected a reference to the dlink as dlink$name to name.dlink
- Changed four occurrences of `F` to `FALSE` in print.summary.dglm.

28 Nov 2006: dglm 1.3
  
Trying again to fix same error as previously. Adding a new input,
etastart=NULL.

25 Sep 2006: dglm 1.2

Fixed an error identified by Sonja Kuhnt for Gaussian models:
Error in eval(expr, envir, enclos) : object "etastart" not found.

05 Sep 2006: dglm 1.1

Fixed error in convergence test identified by Cristiano Amancio for
binomial data. Also fixed warning message "NaNs produced in: log(x)".

15 Sep 2005: dglm 1.0

The function was ported by Peter Dunn to R and submitted to CRAN as
an R package. Peter Dunn registered as package maintainer.

22 Oct 1999

This package was originally a single function, written by
Gordon Smyth in the S language, available from
http://www.statsci.org/s/dglm.html.
The original function is still available from
https://gksmyth.github.io/s/dglm.html.