Detail publikace

Segmentation Based Testing of Co-movement Significance

Originální název

Segmentation Based Testing of Co-movement Significance

Anglický název

Segmentation Based Testing of Co-movement Significance

Jazyk

en

Originální abstrakt

The paper is focused on the significance testing of the time-frequency co-movement measure on the segmentation bases. Investigating the test of the power wavelet cross-spectrum we have some standard assumptions: i.e two independent Gaussian white noise inputs. Then, with the use of the Bessel function, we can test whether the values of power wavelet cross-spectrum are significant with respect to the variance of each input time series. Our paper investigate the case when an input data have heteroscedastic character. Thus we propose firstly segmentation of the data sample according to the variances of input time series. Secondly, we propose an identification significant power wavelet cross-spectrum values in each segment via Ge test. The results with and without segmentation are compared. Our experiment is performed on simulated and real data. The results shows, that segmentation based testing for the heteroscedastic data provides more precise results.

Anglický abstrakt

The paper is focused on the significance testing of the time-frequency co-movement measure on the segmentation bases. Investigating the test of the power wavelet cross-spectrum we have some standard assumptions: i.e two independent Gaussian white noise inputs. Then, with the use of the Bessel function, we can test whether the values of power wavelet cross-spectrum are significant with respect to the variance of each input time series. Our paper investigate the case when an input data have heteroscedastic character. Thus we propose firstly segmentation of the data sample according to the variances of input time series. Secondly, we propose an identification significant power wavelet cross-spectrum values in each segment via Ge test. The results with and without segmentation are compared. Our experiment is performed on simulated and real data. The results shows, that segmentation based testing for the heteroscedastic data provides more precise results.

BibTex


@inproceedings{BUT150980,
  author="Eva {Klejmová} and Tobiáš {Malach} and Jitka {Poměnková}",
  title="Segmentation Based Testing of Co-movement Significance",
  annote="The paper is focused on the significance testing of the time-frequency co-movement measure on the segmentation bases. Investigating the test of the power wavelet cross-spectrum we have some standard assumptions: i.e two independent Gaussian white noise inputs. Then, with the use of the Bessel function, we can test whether the values of power wavelet cross-spectrum are significant with respect to the variance of each input time series. Our paper investigate the case when an input data have heteroscedastic character. Thus we propose firstly segmentation of the data sample according to the variances of input time series. Secondly, we propose an identification significant power wavelet cross-spectrum values in each segment via Ge test. The results with and without segmentation are compared. Our experiment is performed on simulated and real data. The results shows, that segmentation based testing for the heteroscedastic data provides more precise results.
",
  booktitle="Proceedings of the 25th International Conference on Systems, Signals and Image Processing",
  chapter="150980",
  doi="10.1109/IWSSIP.2018.8439304",
  howpublished="online",
  year="2018",
  month="june",
  pages="1--5",
  type="conference paper"
}