- Added
include_probs argument to
cbc_design(), which includes predicted choice probabilities
in the returned design data frame if include_probs = TRUE.
Defaults to FALSE.
- Major overhaul of the package with breaking changes.
- New function, `cbc_priors()``. This allows users to specify a set of
priors according to a wide variety of model specifications, including
random parameters (with or without correlated heterogeneity),
interactions, and “no choice” options. These priors can then be used to
create designs and simulate choices.
- Coefficients for levels of an attribute in
cbc_priors()
can be named vectors, addressing #24.
- Major overhaul of the
cbc_design() function, with
entirely new algorithms for searching for designs
- One is “random”, three are frequency-based (“greedy”) algorithms,
and three more are d-error minimizing algorithms.
- Old methods removed:
"full", "orthogonal",
"dopt", "CEA", and "Modfed"
- Bayesian D-efficient designs are now created based on the priors
provided. With random parameters in the priors, a Bayesian D-efficient
design will be created.
- New support for removing dominant alternatives from designs.
- New support for randomizing the order of questions and alternatives
across respondents, addresses #29.
- New
cbc_inspect() function for comprehensively
inspecting designs.
- New
cbc_compare() function for comparing designs.
- New functionality in
cbc_power() for computing
visualizing power analyses.
- Bug fix in checking input settings (#34)
- Patch to fix a joining issue in the
join_profiles()
function (#27)
- Further revisions to the
method argument in the
cbc_design() function.
- Added the
"random" and "dopt"
methods.
- Added restrictions so that orthogonal designs cannot use the
label argument or restricted profile sets (as either of
these would result in a non-orthogonal design).
- Adjustments made to the
method argument in the
cbc_design() function in preparation for potentially adding
new design methods.
- Added the
"orthogonal" option for generating orthogonal
designs.
- Another small bug fix in
cbc_design() related to #16
where factor level ordering for categorical variables were being
mis-ordered.
- Updated how the
method argument is handled by default
in cbc_design() to be more flexible (anticipating other
methods in the future).
- Added
keep_db_error arg to
cbc_design().
- Bug fix in
cbc_design() where factor level ordering for
categorical variables were being mis-ordered.
- Added additional input check for appropriate
priors in
cbc_design().
- Modify how restrictions are defined in the
cbc_restrict() function to allow users to provide
expressions.
- Add
cbc_restrict() function to improve UI for adding
restrictions to profiles.
- Remove previous approach to including restrictions in
cbc_profiles().
- Add new test cases
- Bug fix: modify code in
cbc_design() to avoid duplicate
choice sets for the same respondents; addresses #7.
- Bug fix: modify code in
cbc_design() to allow Bayesian
D-efficient designs with restricted profile sets; addresses #10 and
#9.
- Added a startup message when the package is loaded.
- Updates for compatibility with logitr version 1.0.1.
- Updated DESCRIPTION and CITATION to remove redundancy in title.
- Updated documentation of returned values in several functions.
- Added initial integration with {idefix} packages for Bayesian
D-efficient designs
- Updates for compatibility with logitr version 0.8.0.
- Updates for compatibility with logitr version 0.7.0.
- Modified the argument of
cbc_profiles() to
... so that the user no longer needs to create a separate
list to define the attributes and levels.
- Modified the arguments for the
randN() and
randLN() functions to mean and
sd.
- Improved printing of counts in
cbc_balance() and
cbc_overlap().
- Updated names of random parameter models to match that of future
logitr v0.6.0.
- Updated documentation and examples for all functions.
- Adding piping example to readme.
- Added support for conditional levels in
cbc_profiles()
- Added a
NEWS.md file to track changes to the
package.