An R package for estimating conditional multivariate reference regions.
The reference region is non parametrically estimated using a kernel density estimator.
Covariates effects on the multivariate response means vector and variance-covariance
matrix, thus on the region shape, are estimated by flexible additive predictors.
Continuous covariates non linear effects might be estimated using penalized splines smoothers.
Confidence intervals for the covariates estimated effects might be derived from
bootstrap resampling. Kernel density bandwidth can be estimated with different methods, including
a method that optimize the region coverage. Numerical, and graphical, summaries
can be obtained by the user in order to evaluate reference region performance with real data.
Full mathematical details can be found in