Differential interval-wise testing for local inference in Sobolev spaces

We present a local non-parametric inferential technique - namely, the differential interval-wise testing, or D-IWT - able to test the distributional equality of two samples of functional data embedded in Sobolev spaces. D-IWT can impute differences between the two samples to specific parts of the domain and to specific orders of differentiation. The proposed technique is applied to the functional data analysis of a data set of tongue profiles.