NEWS

logistf 1.26.0 (2023-08-18)

forward() and backward() now require the dataset as an argument.
Using forward() and backward() to select among factor variables is no longer supported.

Fixed memory bug with large data sets in flic() (#45).
Added support for emmeans: emmeans::emmeans() now works with logistf objects (#51).

Added a NEWS.md file to track changes to the package.
Added terms() to extract terms objects from logistf, flic and flac objects (#32).
logistf() now issues warnings in case of non-convergence.
logistf() more carefully checks its response argument to ensure that they're specified correctly.
logistf() has gained one new control parameter, modcontrol, that controls additional parameters for fitting, e.g. degree of penalization in the fitting process. Default is logistf.mod.control().
Fixed a bug in backward() when data is called data (#34) and when function is called with factors as covariates (#34).
Updated predict() function for easier handling.
logistf(), flic() and flac() are now compatible with termplot() because of the added prediction type terms in predict.logistf(), predict.flic() and predict.flac().
Added a new fitting method: Iteratively reweighted least squares (IRLS).
Added na.action option to logistf() and predict.logistf().
Updated flic() and flac() to be called with interactions in formula.