search

functionaldata / tPACE

Testing version of fdapace
Other
31 stars 22 forks source link

Check correlation of FPCA output #61

Closed MichaelJStein closed 2 years ago

MichaelJStein commented 2 years ago

Hi,

is it possible to check if the FPCA output is uncorrelated? The eigenfunctions should be orthonormal by definition but I would like to know if there is a function to confirm this. Maybe CreateCovPlot is what I am looking for?

Thank you very much!

Best, Michael

aegajardo commented 2 years ago

Hi,

For sparse designs, the estimated FPC scores are correlated. The estimated eigenfunctions should be orthonormal and this can be checked numerically for example using Riemann sum approximations. The function 'trapzRcpp' in fdapace implements the trapezoidal rule for numerical integration. The 'CreateCovPlot' function displays the estimated covariance or correlation surface of the underlying process.