NEWS

lddmm 1.0.0.9000

This major release introduces an improved sampler.

Every participant has one random effect curve for correct responses and one for incorrect responses (random_effects = "correct-incorrect", the new default). A single curve for both is available with random_effects = "shared".
The maximum number of clusters at each block defaults to min(K^2, 2K) (8 for 4 categories) and can be set with z_max.
Spline coefficients that start a curve have the normal prior of article eq. (9), with mean and variance set from the data (hypers$mu_beta0_*, hypers$sigma2_beta0_*).
The smoothness of the fixed effects has a half-Cauchy prior on its standard deviation (scale hypers$s_sigma_*). The upper limits on the variance parameters are documented hyperparameters (hypers$sigma2_1_max, hypers$sigma2_us_max, hypers$sigma2_ua_max).

A single implementation serves all boundary models, and the cluster-label and random-effect updates run in C++. Fits are about 12-21 times faster on the tone learning data.
The cluster labels are updated with an exact locally informed Hamming ball sampler, and the cluster-specific spline coefficients with proposals built from the prior of article eq. (9).
New joint moves improve mixing: a translation between fixed and random effects, a rescaling of the random effects together with their variances, and shifts of the offsets together with the drift and boundary levels.
Coefficients of unused clusters are drawn close to those of the clusters in use (hypers$pseudo_sd), so that curves move between clusters more easily.
Random effects start from each participant's moment estimates, and proposal scales are tuned during burn-in only, in batches of 50 iterations.
Response-time distributions are evaluated on the log scale, which keeps them accurate for slow responses.

New plot_predictive() for posterior predictive checks of the decisions and the response times.
Two worked examples are installed with the package (system.file("examples", package = "lddmm")): toy_example.R analyses the bundled data, and toy_simulation.R simulates data with known parameters and compares the posterior with the truth. The README and the vignette walk through fitting, posterior predictive checks and convergence diagnostics.
New arguments save_z (store cluster labels), save_loglik, informed, verbose and debug_checks; the fit also reports acceptance rates and the settings used.
LDDMM() checks its input and reports problems with the data.
log_likelihood_ind() returns a plain vector, and compute_WAIC() is computed on the log scale.
plot_accuracy() works without attaching dplyr.
The package no longer depends on gtools, LaplacesDemon, plyr, RColorBrewer, RcppProgress and rgen.
Unit tests, and validation scripts in tests/validation/ (a Geweke joint distribution test, invariance tests of individual updates, and a simulate-and-recover study).
The documentation states that response times are in seconds.

Fixed a bug in the version boundaries = "fixed".
Added the function input cluster to determine whether to use the clustering model or the unrestricted model.

Added the option for constant and fixed boundaries over time (boundaries = "fixed-constant").

Now the option for constant boundaries over time (boundaries = "constant") varies across input tone and decision. The same is true for the random effects.
Now the option for fixed boundaries (boundaries = "fixed") varies across time and is fixed across input tone s. The same is true for the random effects.

Now the option for constant boundaries over time (boundaries = "constant") includes same boundaries per each input tone, and varies across possible decisions
Added a function to calculate the Watanabe-Akaike information criterion to compare model performance

Now the boundary parameter in the model can either be unrestricted (boundaries = "flexible"), constant over time (boundaries = "constant") or fixed at the same value across predictors (boundaries = "fixed")

Added the DOI for the article reference
Clarified that response times should be the log transformed of time (in milliseconds) for numerical stability