KSLCore/ksl.utilities.distributions.fitting

Package-level declarations

Types

class ContinuousCDFGoodnessOfFit(data: DoubleArray, val distribution: ContinuousDistributionIfc, val numEstimatedParameters: Int = 1, breakPoints: DoubleArray = suggestBreakPoints(data, distribution)) : DistributionGOF

DiscreteDistributionGOFIfc

interface DiscreteDistributionGOFIfc : DistributionGOFIfc

DiscretePMFGoodnessOfFit

open class DiscretePMFGoodnessOfFit(data: DoubleArray, val distribution: DiscretePMFInRangeDistributionIfc, val numEstimatedParameters: Int = 1, breakPoints: DoubleArray = PMFModeler.makeZeroToInfinityBreakPoints(data.size, distribution)) : DistributionGOF, DiscreteDistributionGOFIfc

DistributionGOF

abstract class DistributionGOF(data: DoubleArray, val numEstimatedParameters: Int = 1, breakPoints: DoubleArray) : DistributionGOFIfc

DistributionGOFIfc

interface DistributionGOFIfc

EstimationResult

class EstimationResult(val originalData: DoubleArray, var statistics: StatisticIfc, var shiftedData: ShiftedData? = null, val parameters: RVParameters? = null, var message: String? = null, var success: Boolean, val estimator: MVBSEstimatorIfc)

A data class to hold information from a parameter estimation algorithm. In general the algorithm may fail due to data or numerical computation issues. The parameters may be null because of such issues; however, there may be cases where the parameters are produced but the algorithm still considers the process a failure as indicated in the success field. The string message allows a general diagnostic explanation of success, failure, or other information about the estimation process. In the case of uni-variate distributions, there may be a shift parameter estimated on shiftedData in order to handle data that has a lower range of domain that does not match well with the distribution. The algorithm may compute statistics on the supplied data.

EvaluationMethod

enum EvaluationMethod : Enum<EvaluationMethod>

Can be used to indicate if the recommended distribution should be based on the MODA scoring model or the first rank frequency across metrics.

PDFModeler

class PDFModeler(observations: DoubleArray, scoringModels: Set<PDFScoringModel> = defaultScoringModels)

PDFModelingResults

data class PDFModelingResults(val estimationResults: List<EstimationResult>, val scoringResults: List<ScoringResult>, val evaluationModel: AdditiveMODAModel)

Holds all the results from the PDF modeling process.

PMFModeler

class PMFModeler(data: IntArray)

PoissonGoodnessOfFit

class PoissonGoodnessOfFit(data: DoubleArray, mean: Double, val numEstimatedParameters: Int = 1, breakPoints: DoubleArray = PMFModeler.makeZeroToInfinityBreakPoints(data.size, Poisson(mean))) : DiscretePMFGoodnessOfFit

ScoringResult

data class ScoringResult(val name: String, val distribution: ContinuousDistributionIfc, val estimationResult: EstimationResult, val rvType: RVParametersTypeIfc, val scores: List<Score>) : Comparable<ScoringResult>

The natural ordering is descending by weighted value. That is scoring results with higher weighted value are considered better (more value is better). The weighted value will be a number within 0,1. Thus, a natural sort will cause elements with higher value to be first in the list. If there are no values, then the weighted value will be zero. The weighting is determined by the scoring method.

ShiftedData

data class ShiftedData(val shift: Double, val shiftedData: DoubleArray)

Functions

fun main()

fun tesNegBinomial()

fun testBinomial()

fun testPoisson()