New dataset vwp_sim: a simulated visual world eyetracking experiment with known main and interaction effects, now used in the README.
Better detection of Julia executable
julia_setup_ok() no longer starts a Julia session. JuliaConnectoR::juliaSetupOk() starts one as of JuliaConnectoR 1.1.6, which left connections open in examples.
clusterpermute() is faster because the permuted timewise statistics now stay in Julia, and only the null cluster-mass distribution is returned to R. Results are identical to the piecemeal workflow.
Faster permutation of predictors (up to ~2x faster clusterpermute() for GLMs). Results are identical to previous versions under the same RNG state.
permute_by_predictor() and permute_timewise_statistics() give an informative error when a predictor is not constant within the units being shuffled (participants, or trials within participants).
Fixed the predictors argument of permute_timewise_statistics() and clusterpermute(), which previously always errored.
extract_null_cluster_dists() no longer drops simulations whose statistics mix Inf and -Inf.
Cluster data frames now have a consistent column order (cluster_id, cluster_start, cluster_end, statistic) regardless of whether the first predictor or simulation contains a cluster.
Fixed crash in mixed-effects models when the response is constant at a time point.
Fixed cluster-mass statistics being returned as a list when a cluster spans a time point with a constant response.
{JuliaFormulae} package for parsing R formula into Julia formulaNo user-visible changes
Fixes for CRAN:
Project.toml to avoid version conflicts.jlmerclusterperm() exposes a cache_dir argument for manually specifying the cache directory. This was added largely for the convenience of testing. The default value of cache_dir = NULL preserves old behavior.Fixes for CRAN:
tempdir() for the purposes of running examples and tests.More informative warnings when constructing a spec object
Fixed minor bugs when interfacing with Julia
Fixed bug with reporting convergence warnings when only one is encountered
Minor breaking change:
permute_timewise_statistics()Fixes for CRAN:
R_user_dir() from {backports} which was missing previouslyFixed bug where add1 argument to walk_threshold_steps() was not being passed down properly
R < 4.0 falls back to using {backports} for R_user_dir()
Fixes for CRAN:
No user-visible changes
Fixes for CRAN:
No user-visible changes
Restructured internal Julia code into a module (JlmerClusterPerm.jl)
Upon activating project, Manifest.toml is cached to tools::R_user_dir() to speed up pre-compilation.
Fixes for CRAN:
jlmerclusterperm_setup() now works for rc and alpha builds of JuliaFixes for CRAN:
jlmerclusterperm_setup() now exits early if Julia version requirement (>=1.8) is not met.
Fixes for CRAN:
threshold_steps argument of walk_threshold_steps() is renamed to steps.get_rng_seed() and set_rng_seed()jlmerclusterperm_setup() now echos Pkg.instantiate() to print precompilation information upon the first setup callAdded vignettes. Significant usability improvements
Initial release