Skip to main content
Fig. 1 | BioData Mining

Fig. 1

From: Uncovering correlated variability in epigenomic datasets using the Karhunen-Loeve transform

Fig. 1

Basic procedure of the functional data analysis. (a) Ten observations of a function, (b) Estimated eigenfunctions; (c) Plot of data in the system of first, second, and third functional principal component scores. Because the eigenfunctions are orthonormal, each dominates others in some subintervals; these subintervals define the occurence of a signal corresponding to each eigenfunction. In (a) selected profiles are marked as green, red, and blue; their order in the subinterval around 0 dominated by first eigenfunction is the same as the order along the first FPCA axis in (c). The same holds approximately (due to loss of information in the two-dimensional graphs) for other functional principal component scores corresponding to eigenfunctions drawn in (b)

Back to article page