kerdec_dens2D: Bivariate Kernel Deconvolution Density Estimation
In gbasulto/kerdec: Kernel deconvolution methods

Description Usage Arguments Details Value Author(s)

This function provides a bandwidth for kernel denvolvolution density estimator. This works for several deconvolution scenarios, specifically, when the error distribution is known, when a sample of pure errors is available instead and when a contaminated sample is available as panel data.

kerdec_dens2D(smp1, smp2, method = c("CV", "NR")[1], kernel = "flat",
  lower = NULL, upper = NULL, x_eval = NULL, h = NULL, h0 = NULL,
  error_smp1 = NULL, error_smp2 = NULL, error_dist = "None",
  error_scale_par1 = NULL, error_scale_par2 = NULL, resolution = 128,
  error_proc = "all", panel_proc = "keep_first",
  truncation_bound = NULL, bw_interval = NULL)

`smp1`	First coordinate of the sample. It is either a vector of size n or a nxl matrix (for repeated observations; l per each individual).
`smp2`	Second coordinate of the sample. It is either a vector of size n or a nxl matrix (for repeated observations; l per each individual).
`method`	Method used to select the bandwidth. The currently available methods are "CV" (cross-validation) and "NR" (normal reference). Cross-validation is the default.
`kernel`	Kernel whose Fourier transform has bounded support. There are currently 5 kernel programmed: "sinc", "triangular", "triw", "tric" and "flat". See vignette for further details and examples. Flat-top kernel is the default.
`lower`	Lower limit where the density will be evaluated.
`upper`	Upper limit where the density will be evaluated.
`x_eval`	Not yet activated. grid of values where the density will be evaluated. If it is given, parameters 'lower' and 'upper' will be omitted.
`h`	Bandwidth parameter which is only required if method = NULL.
`h0`	Optional argument used as initial value to look for the optimal value.
`error_smp1`	Optional vector with the first coordinate of errors. It is necessary to approximate the error distribution if it is unknown.
`error_smp2`	Optional vector with the second coordinate of errors. It is necessary to approximate the error distribution if it is unknown.
`error_dist`	Three possible values are accepted. c("Normal", "Laplace", "None").
`error_scale_par1`	Scale parameter matching the standard deviation of the first coordinate of the error. It is NULL by default and it is required if (and only if) error_dist is normal or Laplace and no sample of error is provided nor contaminated sample comes in panel structure.
`error_scale_par2`	Scale parameter matching the standard deviation of the second coordinate of the error. It is NULL by default and it is required if (and only if) error_dist is normal or Laplace and no sample of error is provided nor contaminated sample comes in panel structure.
`resolution`	Number of points to approximate integral in inversion formula, also to estimate the density (if grid was not given).
`error_proc`	This is required only for panel data structure. It refers to the way errors are processed. See `process_differences` for further details.
`panel_proc`	It defined what will be the contaminated sample (only) for panel data. "keep_first" will use the first column as sample while "take_aver" will take the average of contaminated samples per individual.
`truncation_bound`	It truncates the integrand in deconvolution formula when the denominator is smaller than this bound.
`bw_interval`	Do not modify it.