Infrastructure for extended formulas with multiple parts on the
right-hand side and/or multiple responses on the left-hand side
Changes in Version 1.2-3
o Extended processing of formulas with one or more '.' on the right-hand side: In addition to "separate" and "sequential" processing, there is now dot="previous" which resolves the '.' relative to the previous right-hand side part.
o model.part() failed in case of variables/terms with very long names. Now deparse(..., width.cutoff = 500) is used to support very long variable names as well.
Changes in Version 1.2-2
o Enhance Formula() so that if a "Formula" is supplied it is simply returned unchanged (rather than throwing a warning).
o Enhance update() method for "Formula" object so that also the "new" formula can be a "Formula" object.
Changes in Version 1.2-1
o Bug fix for formulas with transformed variables on the left-hand side (e.g., cbind(), log(), or Surv()) and one ore more '.' on the right-hand side. The terms() and hence the model.frame() now work smoothly. When using model.part() the same Formula (plus 'dot' argument) has to be supplied when preparing the model.frame() and the model.part().
Changes in Version 1.2-0
o Extended processing of formulas with one or more '.' on the right-hand side.
Changes in Version 1.1-2
o Added a str() method.
Changes in Version 1.1-1
o The CITATION was incorrect and is fixed now.
Changes in Version 1.1-0
o All methods returning "formula" or "Formula" objects now preserve the environment of the originally supplied object by default.
o The default and formula methods of as.Formula() methods now also take an env argument.
Changes in Version 1.0-1
o Added a terms = FALSE argument to model.part() method for "Formula" objects. This can be leveraged when processing multiple offsets. For example for y ~ x + offset(o1) | z + offset(o2). See ?model.part for a worked example.
Changes in Version 1.0-0
o package now published in Journal of Statistical Software, see http://www.jstatsoft.org/v34/i01/ and citation("Formula") within R.
o added an all.equal() method for "Formula" objects that produces more intelligible output in case the result is not TRUE.
o fixed an error of the update() method for "Formula" objects without LHS.
Changes in Version 0.2-0
o Major revision (not fully backward compatible) to enable support for multiple responses and multiple parts on the right-hand side such as y1 | y2 ~ x y1 + y2 ~ x1 + x2 | z1 y ~ u1 + u2 | v1 | x1 + x2 and combinations of these.
o Formula objects now consist of the original formula plus two attributes "lhs" and "rhs" that contain the parts of the decomposed left- and right-hand side, respectively.
o Most methods take arguments "lhs" and "rhs" which allow selection of the desired parts on the left- and right-hand side respectively, e.g., in a model.frame() or model.matrix().
o The previous arguments response = TRUE/FALSE and part = "first"/"second"/"both" were not flexible enough anymore and have been deprecated. Use the streamlined lhs/rhs arguments instead.
o vignette("Formula", package = "Formula") illustrates usage of the tools provided by the package and explains the ideas underlying its implementation.
Changes in Version 0.1-3
o added CITATION file.
o added a has.intercept function with methods for formula and Formula objects.
o change the default value for model.frame.Formula: "both" for a two-part formula and "first" otherwise.
Changes in Version 0.1-2
o Bug fix: as.Formula() failed for very long formulas.
Changes in Version 0.1-1
o First CRAN release of package "Formula" for extended formula processing. This package is still under development and the interface might change in future versions.
o Currently, this enables processing of formulas such as y ~ x1 + x2 | z1 + z2 + z3 with two parts on the right hand side.
o Generalization to further formulas are planned (i.e., not implemented yet), e.g., y1 + y2 ~ x1 + x2 + x3 y ~ x1 + x2 | u1 | v1 + v2 + v3 | ... i.e., multiple responses and multi-part formulas.