- Enhance internal function of
calc_sdt(). This function could be exported in future.
- Fixed an issue of binding data.frames by position, which is now corrected by matching columns by name.
- Removed
wrangle_data()and now there istarflow.iquizoo::parse_data()to replace it (#86).
- Now case conversion is moved from
wrangle_data()topreproc_data()(#85).
- Added
memory_scoretodrm()which integrates both true memory and false memory.
- Added
"cutoff"method in response time outlier detection, which is used insrt()now. - Added
"transform"method in response time outlier detection, which was stated to be used but the code was not right. Now it is used as default if not specified.
- Let internal function
calc_sdt()throw error when type column is not valid. - Enhanced code quality of internal functions.
- Handle duplicated lower-case column names by skipping conversion.
- Fixed observation order issue introduced by
calc_indices().
- Added
wrangle_data()andpreproc_data()functions, which were previously in tarflow.iquizoo package.
- Fixed an edge case when
fit_numerosity()will stuck in infinite loop.
- Enhance code quality of internal functions.
- Ensure the number of reversals in
calc_staircase_wetherill()to be even. Odd number of reversals will be automatically adjusted to even number by removing the first reversal from the condition with larger number of reversals (#75).
- Added
condstairs()to analyze data from self-adaptive filtering task. - Added
capacity()to analyze data from visual working memory capacity task. - Added
fname()to analyze data from face-name associative task. - Added
ant_orient()andant_alert()to analyze data from split version of attention network tests (ANT). - Added support for extra checking for valid rows to include in
countcorrect(). - Added support for restoring
"outcome"column from"acc"column inspan(). This might need further configuration of"outcome_rule"(default to 1, means no more error count of 1) in argument.extra.
- Fixed the random seed in
fit_numerosity(). Future work might be expanded to support setting this (rather easy but not so important now).
- After {tidyselect} 1.2.0 and {dplyr} 1.1.0, many old usage of {tidyverse} functions will throw warnings. These warnings are now taken care of.
- Fixed a bug in
span()when correctness column hasNAvalues not in all rows, in which case the function would not calculate partical credit scores for all cases.
- Reverted one index (
pc) ofcpt()to its originalnc.
- Added
.byargument back to all functions for speed improvement.
- Ensured all the functions return data with the same
classas inputdata. This is somewhat so-called class invariant.
- Added
rapm()for Raven's Advanced Progressive Matrices test. - Exported
fit_numerosity(). - Added basic support for indices combining speed and accuracy performances in
calc_spd_acc(). - Supported new version of Racer test in
racer(). - Added index of percent correct for learning phase for reinforcement test in
reinf().
- Fixed bug of switch cost for percent correct (#68).
- Fixed docstring of
span()(#65).
- Correct spellings.
- Now the
.byargument from all functions is removed. The decision to include that grouping argument is now considered a redundant feature.
- Added new function
racer()to support NeuroRacer tests.
- Added new function
reinf()to support for reinforcement learning tests. - Support threshold calculation in
staircase()with inflection point.
- Fixed issues for
synwin()caused by a new design of the test. - Now response times are corrected for no-response trials in stop signal test.
- Added new function
dualnback()to support for dual n-back tests. - Added new function
staircase()to support for tests shipped with staircase method. - Added new function
synwin()to support modified SynWin test. - Now
span()can deal with location memory test. - Supported new design of london tower test.
- Added partical credit score for simple or complex span tests (#36).
- Changed algorithm to get stop signal reaction time for Stop Signal Paradigm based on the recent eLife publication (#56).
- Added support for ncorrect and nerror input for
countcorrect()andcountcorrect2(). - Enhanced reaction time outlier detection (#61).
- Enhanced fitting of weber fraction for numerosity task through trial-based maximum likelihood estimation.
- Now
options()support is removed, but accept two additional arguments.inputand.extra.
- Remove
wherefrom global variables.
- Removed support for custom-defined json configuration file.
- Added support to configure function behavior in
options(), for now including input variable names in data and specific values for certain variables. - Now
preproc()is removed because it is believed there is no place for it after this refactoring.
- Remove several unused package from dependency.
- Changed argument names of
prep_funandbyto.fnand.byrespectively. See here for the reasons. - And
vars_inputis also renamed as.input.
- Remove "Datasets" section from pkgdown site.
- Now internal data moved to {data.iquizoo} package.
- Now
preproc()accepts custom-defined functions and custom-defined configurations (input and output variable names) in a custom.jsonfile.
- Now the package is named as "preproc.iquizoo".
game_infodata is moved to {data.iquizoo} package.
- For integrating "{dm}" package, we moved
wrangle_data()back to {tarflow.iquizoo}, and then renamedpreproc_data()topreproc().
- Added more games into the
game_infodata.
- Added a new function
wrange_data()(originally in {tarflow.iquizoo}). Now it removes duplicates from data. - Added two more indices in
refframe()of distance error aggregating both conditions.
- Support an edge case for
span()when number of correct responses can be recovered from columns"stim"and"resp". Use it with care, for it does take order into consideration!
- Support input of cancellation games data for
cpt().
- Now cancellation games use
cpt()preprocessing function.
- Changed the behavior of
preproc_data()to make sure it restores all the attributes from inputdata.
- Now pipe and rlang helpers are not re-exported from this package.
- Add a subtle new feature that now more keywords are supported in
nback(). Specifically, you can now use"filler"and"target"keywords. - Add another subtle new feature so that you can ignore the case of character vectors. For example,
"filler"and"Filler"are both okay for now, but only"Filler"is errorproof for previous versions.
- Added more games in the data
game_info, mostly are questionnaires (#27, thanks to @Blockhead-yj). - Fixed case issues of data name. Now lowercase is used thoroughly for data names (note it is not snake_case, for readability is not for consideration).
- Add support for group-wise calculation, and this is implemented by using a new "entry-point" function
preproc_data()with an input argument ofby(#23, #25).
- Use log-linear correction for extreme proportions in calculation of d' based on signal detection theory (#16).
- Add more games to
game_infodata.
- Completely removed
conflict(), which is defunct for many versions. - Used the 3rd version test framework of testthat package, especially takes advantage of the snapshot test (or golden test) to simplify all our tests.
- Unified the abnormal output by setting all of the
NAs asNA_real. - Unified the response metrics checking. Now many games need to have a minimal valid response rate of 80% and a minimal accuracy rate, too. Use
?tarflow.iquizoo:::check_resp_metricto see details. Note this function is not exported for now.
- Add external data
game_info, which stores games information of preprocessing functions. Run?game_infoin R to read more details.
- Fixed a bug occured in function
london()when user did not respond and the data still scored the user as correct. - Added a response rate check for
london(). Now the score is normal only when the response rate is no less than 80%.
- Just as
complexswitch(), nowswitchcost()will also return abnormal result (allNAs) if at least one block has no responses.
- Fix a bug occured in function
multisense()when users did not respond to sound trials.
- Fix a bug occured in function
wxpred()after fillingNAvalues toname_block(mostly is justBlock) column. - Fix a bug occured in function
driving()when minus signs, which are confused with hyphens, are logged into data.
- Now
complexswitch()will return abnormal result (allNAs) if at least one block has no responses.
- Rename package from "preprocr2" to "preproc.iquizoo". This is recommended because all the work is related to www.iquizoo.com.
- Added a utility function
calc_sdt()to calculate signal detection theory related indices. - Fixed an issue related to perfect accuracy in
drm().
- Added a
NEWS.mdfile to track changes to the package. - Added
pc_allindex, meaning overal percent of correct, towxpredfunction.