2021-10-08
* version 1.2.0 to 1.2.1
- documentation house keeping
* added dag.sim2
- extends the functionality of dag.sim
- binary nodes can be simulated based on linear risk difference model
- for both this new specification and the previously available simulations
  based on the logistic model, the success probability can now be specified
  either for the overall cohort (as before) or for the reference category
  (i.e., where the summed up weighted predictors equal 0)
- most dag.sim calls should be compatible with dag.sim2, but both functions
  will be available for the time being

2021-04-24
* version 1.1.3 to 1.2.0
- added generic print, plot, and summary functions
- added a function to create dagitty syntax from a dagR DAG

2014-01-08
* version 1.1.2 to 1.1.3
* dag.sim

2012-11-04
* version 1.1.1 to 1.1.2
* dag.draw
- new option to allow NOT drawing the X->Y arc, or drawing it w/out "?"
* changes to multiple functions to allow custom node symbols
- no special formatting has been programmed, i.e. the string length is ignored
- affected functions: dag.draw, dag.init, dag.legend, dag.letter, write.paths, summary_dagRdag
* dag.letter
- added note on adjustment marking by bar/underline (not for unknowns) in docu.
* new function: demo.dag7
- same as demo.dag3, but with alternative node symbols

2010-06-14
* version 1.0.1 to 1.1.1
* removed empty sections from .Rd files
* dag.init
- now adds version of dagR to the DAG
- now assigns class "dagRdag" 
- elaborated documentation regarding node numbering
* rm.node, add.arc, rm.arc
- now also remove searchType and searchRes
* dag.adjust
- if adjustment set not empty, now also removes searchType and searchRes
- elaborated documentation regarding node numbering
* is.acyclic
- new argument: maxSecs
* new function: viv
* new function: msas
* new function: dag.search
* new function: summary_dagRdag

OPEN ISSUES
* brute.search
- has trouble if there are no eligible adjustment sets to evaluate
  (e.g. only unknown variables, but allow.unknowns=FALSE)
* dag.draw
- for some reason, this sometimes has a problem because components
  of dag$arc strangely become lists instead of numeric
* dag.letter
- unknowns are not shown with adjustment marking
* helper functions should go internal...
* more recently, better algorithms used to identify open backdoor
  paths have been published and should be implemented in dagR
  (or at least improve pertinent interface to dagitty)
* add function to convert dagitty/ggdag DAG objects to dagR objects
* extend simulation capacities to additional distributions (trying not
  to break backwards compatibility?) and
  convenient repeated sampling (output list?)