BSW 0.1.2
New
- Added option conswitchinbsw():
- 1(default): all possible min/max combinations of
predictors (bounds predictions).
- 0: raw design matrix (suitable for risk factor
identification only).
 
- New function obj_value()to compute the log-likelihood
value used in the Armijo line search.
- Implemented a robust bootstrap procedure bootBSW()forbswobjects. The function completes execution even if
individual bootstrap samples fail to converge.
- Implemented variable selection methods
(variable_selection_bsw()), supporting both backward
elimination and forward selection. User can specify significance level
(alpha) and maximum iterations (maxit).
Improved
- Optimized Hessian matrix calculation by replacing explicit loops
with vectorized operations.
- Armijo line search implemented to ensure stable convergence.