Interval-wise testing for functional data

In the framework of null hypothesis significance testing for functional data, we propose a procedure able to select intervals of the domain imputable for the rejection of a null hypothesis. An unadjusted p-value function and an adjusted one are the output of the procedure, namely interval-wise testi...

Full description

Saved in:
Bibliographic Details
Published in:Journal of nonparametric statistics 2017-04, Vol.29 (2), p.407-424
Main Authors: Pini, A., Vantini, S.
Format: Article
Language:English
Subjects:
Adjustment
Benchmarking
canadian temperatures
domain selection
Errors
functional data
Gain
Inference
Intervals
Nonparametric statistics
Null hypothesis
Regression analysis
State of the art
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In the framework of null hypothesis significance testing for functional data, we propose a procedure able to select intervals of the domain imputable for the rejection of a null hypothesis. An unadjusted p-value function and an adjusted one are the output of the procedure, namely interval-wise testing. Depending on the sort and level α of type-I error control, significant intervals can be selected by thresholding the two p-value functions at level α. We prove that the unadjusted (adjusted) p-value function point-wise (interval-wise) controls the probability of type-I error and it is point-wise (interval-wise) consistent. To enlighten the gain in terms of interpretation of the phenomenon under study, we applied the interval-wise testing to the analysis of a benchmark functional data set, i.e. Canadian daily temperatures. The new procedure provides insights that current state-of-the-art procedures do not, supporting similar advantages in the analysis of functional data with less prior knowledge.
ISSN:1048-5252
1029-0311
DOI:10.1080/10485252.2017.1306627