kernopt - A Package for estimating count data distributions with a Discrete Symmetric Optimal Kernel

kernopt:

kernopt is an R package that implements Discrete Symmetric Optimal Kernel for estimating count data distributions, as described by (Senga Kiessé and Durrieu 2024). The nonparametric estimator using the discrete symmetric optimal kernel was illustrated on simulated data sets and a real-word data set included in the package, in comparison with two other discrete symmetric kernels.

Authors:

• Tristan Senga Kiessé, UMR SAS INRAE, Institut Agro
• Gilles Durrieu, Université Bretagne Sud - CNRS UMR 6205, LMBA
• Thomas Fillon, Université Bretagne Sud - CNRS UMR 6205, LMBA & CNRS UMR 6074 IRISA

Installation

You can install the development version of kernopt from GitHub with:

# install.packages("pak")
pak::pak("thomasfillon/kernopt")

Example

This is a basic example which shows how to use the kernopt library to compute the discrete optimal kernel values for some parameters:

library(kernopt)

## Compute the discrete optimal kernel values
k_opt <- discrete_optimal(x = 25, z = 1:50, h = 0.9, k = 20)
print(k_opt)
#>  [1] 0.00000000 0.00000000 0.00000000 0.00000000 0.01871809 0.01956892
#>  [7] 0.02037611 0.02113967 0.02185959 0.02253589 0.02316855 0.02375758
#> [13] 0.02430298 0.02480475 0.02526288 0.02567739 0.02604826 0.02637550
#> [19] 0.02665910 0.02689908 0.02709542 0.02724813 0.02735721 0.02742266
#> [25] 0.02744448 0.02742266 0.02735721 0.02724813 0.02709542 0.02689908
#> [31] 0.02665910 0.02637550 0.02604826 0.02567739 0.02526288 0.02480475
#> [37] 0.02430298 0.02375758 0.02316855 0.02253589 0.02185959 0.02113967
#> [43] 0.02037611 0.01956892 0.01871809 0.00000000 0.00000000 0.00000000
#> [49] 0.00000000 0.00000000

The documentation is available at https://thomasfillon.github.io/kernopt/.

References

Senga Kiessé, Tristan, and Gilles Durrieu. 2024. “On a Discrete Symmetric Optimal Associated Kernel for Estimating Count Data Distributions.” Statistics & Probability Letters 208: 110078. https://doi.org/10.1016/j.spl.2024.110078.