NEWS

mfp2 1.0.2

Added spike-at-zero algorithm for modeling semicontinuous covariates.
mfp2(): Introduced zero and catzero options to model only the positive part of a semicontinuous covariate.
mfp2(): family argument now supports GLM family functions (e.g., stats::binomial(link = "probit")) and different link functions.
predict.mfp2(): Fixed bug when using type = "response" with the binomial family.
transform_data_step(): Improved efficiency by transforming a variable only when its fractional polynomial power or spike decision changes.
fracplot(): Extended to allow plotting variables with a spike at zero.

Final regression coefficients are now expressed on the original scale of the data.
predict.mfp2(): Fixed a bug when using type = "terms".
predict.mfp2(): Added the nseq argument.
mfp2(): Fixed an issue where the keep argument was inactive when criterion = "aic" or criterion = "bic".
Fixed a bug in BIC calculation for Cox models: nobs now correctly uses the number of events rather than the total number of observations.
Improved documentation for clarity and added references regarding the effects of influential points in FP functions.
DESCRIPTION: Removed NeedsCompilation field as the package contains no native code.