NEWS

spareg 1.2.0 (2026-09-25)

Updated plot methods to display the original scale for the threshold parameter when plot_along = "nu".
Remove the duplicated warning message in the parallel = TRUE case.
Made the name argument optional in constructor functions.
Extended fast fitting options in spar.cv to include (using new argument precompute_mode):
- Fixing screening indicators and random projections in advance -- precompute_mode = "precompute_all".
- Fixing only the random projections in advance -- precompute_mode = "precompute_rpm".
- Sampling both screening indicators and random projections in each fold of the full cross-validation -- precompute_mode = "full_cv".
Restricted plot.spar.cv, coef.spar.cv, predict.spar.cv to be used only for precompute_mode = "precompute_all".
Improved consistency across the S3 classes randomprojection, screencoef and sparmodel. They all now rely on an update function and a generate function.
Changed thresholding to beta[abs(beta) <= nu] <- 0 instead of beta[abs(beta) < nu] <- 0

Enhance print method with type of measure being used for choosing $M$ and $\nu$
Allow for cases where $p < n$.
Allow specification of possible beta values in simulate_spareg_data().
Changed robust example in vignette to using a poisson() link.
Reproducibility of parallel case is introduced using doRNG.
Added extractor function extract_spareg() to extract fitted values, residuals, and coefficients from a fitted model.
Added extractor function get_model() to extract the best or the 1se model.
Added extractor function get_measure() to extract the table of (cross-) validation measures for the grid of (nu, nummods).
Added class coefspar with print and summary methods to improve the usability of the coef method.
Added extractor functions get_intercept() and get_coef() to extract the intercept and coefficients (non-standardized) from the coefspar objects.
Added avg_type i.e., type of averaging the marginal models to spar(). This argument is used in computing the validation measure.
Removed coef argument from predict method.
Added aggregate = c("mean", "median") argument for predict method.