spima 0.2.0
- Distance function fix (v0.2.0):
spima_cont_distance changed from relative-difference + SD
mixing to inverse-variance weighted Euclidean distance, eliminating
systematic bias in the continuous module.
- Mixed-format support: all 5 input formats (mean+SD,
median+IQR, median+range, mean+range, five-number summary) can be mixed
in a single dataset.
- Forest plot: new
forest.spima() S3
method with log_scale and custom label support. Alias
spima_forest() works when metafor masks the
generic.
- Adaptive particles:
smc_control()
gains n_particles_max; ABC-SMC doubles particle count when
acceptance rate is low.
- Parallel fallback: graceful reversion to sequential
when PSOCK cluster fails.
- Precision weights:
spima_cont_observed_stats now attaches proper SE-based
weights for more accurate distance computation.
- Improved
as.data.frame(): new
probs parameter for custom posterior quantiles.
spima 0.1.0
- Initial release.
- Meta-analysis via ABC-SMC by simulating pseudo-individual data from
published group-level summary statistics.
- Modules: binary, continuous (Gaussian and Gamma, with flexible input
formats), generic effect-size, and interaction analysis.
- C++ accelerated simulation and distance computation.