GetFVE {fdadensity}

Compute Metric-based Fraction of Variance Explained

Description

When FPCA is performed on the log quantile density functions, the fraction of variance explained by the first K components is computed based on the density reconstruction and chosen metric.

Usage

GetFVE(fpcaObj, dmatrix, dSup, metric = "L2", useAlpha = FALSE, alpha = 0.01)

Arguments

fpcaObj

PACE output (FPCA on LQDs)

dmatrix

matrix of original densities measures on grid dSup, rows correspond to individual densities

dSup

support for Density domain - max and min mark the boundary of the support.

metric

metric for measuring variance - 'L2' for Euclidean or 'W' for Wasserstein

useAlpha

should regularisation be performed to densities in dmatrix? This should be set to TRUE if densities were regularised prior to FPCA (default = FALSE)

alpha

scalar to regularise before computing FVE. If useAlpha = TRUE, this should match the value used to regularise prior to FPCA (default = 0.01)

Details

The fraction of variance explained (FVE) by the first K principal components corresponding to the LQD functions is computed by taking the K-dimensional LQD representations, transforming back to densities, and comparing the reconstruction to the original densities using the chosen metric. If densities were regularised prior to transformation and FPCA, the same regularisation parameters should be used here.

Value

FVEvector

References

Functional Data Analysis for Density Functions by Transformation to a Hilbert space, Alexander Petersen and Hans-Georg Mueller, 2016

See Also

RegulariseByAlpha,lqd2quantile

Examples

data(Top50BabyNames)

# Perform Transformation FPCA for male baby name densities
dSup = Top50BabyNames$x
X = FPCAdens(dmatrix = t(Top50BabyNames$dens$male), dSup = dSup, useAlpha = TRUE, 
                   optns = list(dataType = 'Dense', error = FALSE, methodSelectK = 8))

# Compute FVE - must compare to regularized densities 

fveL2 = GetFVE(fpcaObj = X, dmatrix = t(Top50BabyNames$dens$male), dSup = dSup, useAlpha = TRUE)
fveW = GetFVE(fpcaObj = X, dmatrix = t(Top50BabyNames$dens$male), dSup = dSup, 
                   metric = 'W', useAlpha = TRUE)