KernelDensity.FitTransformed Method

Fits a boundary-corrected density in log or logit coordinates.

Definition

Namespace: Numerics.NET.Statistics
Assembly: Numerics.NET (in Numerics.NET.dll) Version: 10.9.0

public static KernelDensityEstimate FitTransformed(
	Vector<double> input,
	KernelDensityTransformation transformation,
	KernelDensityOptions? options = null,
	Vector<double>? weights = null
)

Visual Basic

Public Shared Function FitTransformed ( 
	input As Vector(Of Double),
	transformation As KernelDensityTransformation,
	Optional options As KernelDensityOptions = Nothing,
	Optional weights As Vector(Of Double) = Nothing
) As KernelDensityEstimate

C++/CLI

public:
static KernelDensityEstimate^ FitTransformed(
	Vector<double>^ input, 
	KernelDensityTransformation^ transformation, 
	KernelDensityOptions^ options = nullptr, 
	Vector<double>^ weights = nullptr
)

static member FitTransformed : 
        input : Vector<float> * 
        transformation : KernelDensityTransformation * 
        ?options : KernelDensityOptions * 
        ?weights : Vector<float> 
(* Defaults:
        let _options = defaultArg options null
        let _weights = defaultArg weights null
*)
-> KernelDensityEstimate

Parameters

input Vector<Double>: Nonempty finite observations, strictly inside the transformation bounds, including any zero-weight entries. Observations are snapshotted.
transformation KernelDensityTransformation: The monotone transformation and declared support bounds.
options KernelDensityOptions (Optional): Kernel and bandwidth options in transformed coordinates, or null for Gaussian normal reference.
weights Vector<Double> (Optional): Optional finite nonnegative relative observation masses, paired by position, matching the input length and containing positive mass. Null gives equal mass.

Return Value

KernelDensityEstimate
An immutable transformed fit. Queries and densities use original coordinates; bandwidth metadata is exposed directly in transformed coordinates.

Remarks

The change-of-variable Jacobian is included in density and log-density. Fixed widths, adjustment and bandwidth-selection criteria apply to transformed observations. Endpoint observations are rejected; no arbitrary epsilon offset is applied. Existing unbounded fitting is unchanged.

Exceptions

ArgumentNullException	The input or transformation is null.
ArgumentException	An observation or weight is invalid, automatic selection has insufficient transformed spread, or the selected width is not representable.