NEWS

calmr 0.6.3

Added set_calmr_palette() function to control the colour/fill scales used to plot results (#1).
Fixed bug in make_experiment() that was triggered by empty phases and no miniblocks.
Changed get_timings() to require a specific model name.
Added vignette for TD model.

Aggregation of ANCCR data now ignores time; time entries are averaged.
Added the Temporal Difference model under the name "TD". The model is in an experimental state.
Experiments for time-based models now require a separate list to construct time-based experiences. See get_timings().
Added experiences<-, timings, timings<- methods for CalmrExperiment class.
Revamped plotting functions and parsing functions.
Revamped output names for all models to make them more intelligible.
Fixed a bug related to the aggregation of pools in HDI2020 and HD2022.
Consolidated some man pages.

Added outputs argument to run_experiment(), parse(), and aggregate(), allowing the user to parse/aggregate only some model outputs.
Documentation corrections for CRAN resubmission.

Added dependency on data.table resulting in great speedups for large experiments.
Replaced dependency on cowplot with dependency on patchwork.
Removed dependencies on tibble, dplyr, tidyr, and other packages from the tidyverse.
Removed shiny app from the package.
The previous app is now distributed separately via the calmr.app package available on GitHub.
Test coverage has reached 100%.
The package is now ready for CRAN submission.

Added parallelization and progress bars via future, future.apply, and progressr.
Function calmr_verbosity can set the verbosity of the package.

Implementation of ANCCR (Jeong et al., 2022), the first time-based model included in calmr.
Added parameter distinction between trial-wise and period-wise parameters.
Added internal augmentation of arguments depending on the model.
All trial-based models do not use pre/post distinctions anymore. Using the ">" special character does not affect these models anymore.
The ">" special character is used to specify periods within a trial. For example, "A>B>C" implies A is followed by B which is followed by C. See the using_time_models vignette for additional information.
Named stimuli now support numbers trailing characters (e.g., "(US1)" is valid now.)

Major refactoring of classes and models. This should help development moving forward.
Added several methods for access to CalmrExperiment contents, including c (to bind experiments) results, plot, graph, design, and parameters.
Created CalmrDesign and CalmrResult classes.
Rewrote parsers to be less verbose and to rely less on the tidyverse suite and piping.
Substantially reduced the complexity of make_experiment function (previous make_experiment).
Introduced distinction between stimulus-specific and global parameters.
Parameters are now lists instead of data.frames.
Modified UI for calmr app to include a sidebar.
Simplified the app by removing some of the options.
Nearly duplicated the number of tests.

Added first version of the SOCR model (SM2007) as well as two vignettes explaining the math behind the implementation and some quick simulations.
Documentation progress.

Added multiple models to package and app (RW1972, PKH1982, MAC1975).
Implementation of basic S4 classes for model, experiment, fit, and RSA comparison objects, as well as their methods.
Added genetic algorithms (via GA) for parameter estimation.
Added basic tools to perform representational similarity analysis.
Documentation progress.

heidi is now calmr. The package now aims to maintain several associative learning models and implement tools for their use.
Major overhaul of the training function (train_pav_model). All relevant calculations are now done as a function of all functional stimuli instead of just the US.
Support for the specification of expectation/correction steps within the trial via ">". For example, the trial "A>(US)" will use only A to generate the expectation, but will learn about both stimuli during the correction step.
The previous plotting function for R-values has been revamped to allow both simple and complex versions. The complex version facets r-values on a predictor basis, and uses colour lines for each target.
Bugfix related to stimulus saliencies.