serp: Smooth Effects on Response Penalty for CLM
Implements a regularization method for cumulative link models using the
Smooth-Effect-on-Response Penalty (SERP). This method allows flexible
modeling of ordinal data by enabling a smooth transition from a general
cumulative link model to a simplified version of the same model. As the
tuning parameter increases from zero to infinity, the subject-specific
effects for each variable converge to a single global effect.
The approach addresses common issues in cumulative link models, such as
parameter unidentifiability and numerical instability, by maximizing a
penalized log-likelihood instead of the standard non-penalized version.
Fitting is performed using a modified Newton's method. Additionally, the
package includes various model performance metrics and descriptive tools.
For details on the implemented penalty method, see
Ugba (2021) <doi:10.21105/joss.03705> and
Ugba et al. (2021) <doi:10.3390/stats4030037>.
Documentation:
Downloads:
Reverse dependencies:
Linking:
Please use the canonical form
https://CRAN.R-project.org/package=serp
to link to this page.
[aut, cre, cph]