Forum: open-discussion

Poincare plot confidence [ Reply ]
By: Toru Hitomi on 2024-03-01 13:33

Hi,

I've read source codes for Poincare plot (RHRV, v4.2.7 from https://rdrr.io/cran/RHRV/src/R/poincarePlot.R) and found something unclear to me.

Basically the PoincarePlot function has arguments regarding the confidence level spec (`confidenceEstimation` and `confidence`).
confidenceEstimation is just a boolean to decide if confidence is estimated or not. The thing I want to know is about `confidence`, which seems to specify the confidence level (Default: 0.95).
Specifically, the confidence argument is passed to the computeSD() and then confidenceEllipse() function:

[In PoincarePlot]
SD = computeSD(timeSeries = rrSeries, timeLag = timeLag,
confidenceEstimation = confidenceEstimation,
confidence = confidence)

[In computeSD]
if (confidenceEstimation) {
takens = buildTakens(time.series = timeSeries,
embedding.dim = 2,
time.lag = timeLag)
SD = confidenceEllipse(x = takens[, 1],
y = takens[, 2],
confidence = confidence)
}

[In confidenceEllipse]
confidenceEllipse <- function(x,y, confidence=0.95){
# Get covariance matrix and mean point
covMatrix = cov(cbind(x,y))
# Compute eigenvalues and eigenvectors
eigenComputations = eigen(covMatrix)
eigenvalues = eigenComputations$values
eigenvectors = eigenComputations$vectors
# rotate eigenvectors to obtain "usual" directions
# first eigenvector must have all components > 0
if ( (eigenvectors[1,1]*eigenvectors[1,2]) < 0) {
rotationMatrix = matrix(c(-1,0,0,-1), ncol = 2,
byrow = FALSE)
eigenvectors = rotationMatrix %*% eigenvectors
}

return(list(sd = sqrt(eigenvalues), directions = eigenvectors))
}

However, if you carefully look at the confidenceEllipse function, the function receives confidence level argument but doesn't use it at all.

My questions are:
1. Is this a bug?
2. When you calculate confidence estimation, the argument `confidence` won't be used?

I hope that my description makes sense and welcome any suggestions or comments.

Best regards,
Toru