- The interface of
hermiterhas been simplified. Theupdate_batchmethod has been removed in favor of providing the ability to initialize thehermite_estimatorwith an initial batch of observations. Several internal methods are no longer exported in the interests of simplicity. - The default value of N for univariate estimators has been changed from N = 30 to N = 50.
- Parallel implementation of batch updating using RcppParallel provides significant performance improvements on multicore systems.
- Updated citation information.
- Additional test cases have been added.
- Bug fixes for series acceleration algorithm.
- A method has been added for estimating the Kendall rank correlation coefficient in the bivariate setting.
- The univariate quantile estimation method has been significantly enhanced in accuracy using series acceleration techniques. Series acceleration is enabled by default.
- The univariate pdf and cdf methods have been significantly enhanced in accuracy using series acceleration techniques. Series acceleration is enabled by default.
- The new default method for the univariate quantile estimation method, 'interpolate' is much faster than the alternate method, 'bisection' with nearly the same accuracy.
- Added print and summary methods for both the univariate and bivariate hermite_estimator objects.
- Convenience function added to calculate sums of Hermite functions.
- The vignette
hermiter, namelyvignette("hermiter")has been extended to included examples pertaining to estimation of the Kendall Tau nonparametric correlation coefficient in the bivariate setting.
- Minor update to test cases.
get_coefficientshas been removed as it is redundant.combine_hermitehas been renamed tomerge_hermitefor clarity.combine_pairhas been renamed tomerge_pairfor clarity.hermite_integral_val_quantile_adaphas been renamed tohermite_integral_val_upperfor clarity.
- Bivariate Hermite estimators have been added with methods for estimating bivariate probability density functions and cumulative distribution functions along with Spearman's rank correlation coefficients.
- The bivariate estimators include methods to batch update or sequentially update.
- Methods are also provided to consistently merge bivariate hermite_estimators.
- Convenience methods have been added for calculating normalized Hermite functions, along with upper, lower and full domain integrals of the normalized Hermite functions.
- The vignette
hermiter, namelyvignette("hermiter")has been extended to included examples pertaining to the bivariate Hermite series based estimators.
- The method for merging univariate Hermite series based estimators has been improved, yielding greater accuracy when the hermite_estimators are standardized.
- The method for estimating quantiles with the univariate Hermite series based estimator has been improved and is now consistent with the estimator in the literature.
- Vectorized the univariate quantile estimation method.
- Added further error trapping and other minor enhancements (also for C++ routines).