Cross wavelet analysis: significance testing and pitfalls
By Douglas Maraun, J. Kurths. 2004
139 citations recorded by the source corpus.
Abstract. In this paper, we present a detailed evaluation of cross wavelet analysis of bivariate time series. We develop a statistical test for zero wavelet coherency based on Monte Carlo simulations. If at least one of the two processes considered is Gaussian white noise, an approximative formula for the critical value can be utilized. In a second part, typical pitfalls of wavelet cross spectra and wavelet coherency are discussed. The wavelet cross spectrum appears to be not suitable for significance testing the interrelation between two processes. Instead, one should rather apply wavelet coherency. Furthermore we investigate problems due to multiple testing. Based on these results, we show that coherency between ENSO and NAO is an artefact for most of the time from 1900 to 1995. However, during a distinct period from around 1920 to 1940, significant coherency between the two phenomena occurs.
Subfields: Computer Vision and Pattern Recognition, Economics and Econometrics
View Paper: https://www.semanticscholar.org/paper/9ce3b6ab2ecfb99b46272ed989f9f381e9ee3b92
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computer Vision and Pattern Recognition
Citations in the corpus
- Geophysical Applications of Partial Wavelet Coherence and Multiple Wavelet Coherence (2012, background): “Maraun and Kurths (2004) presented the evolution of cross-wavelet transform, developed a statistical test for wavelet coherence (WTC), 1 and discussed some pitfalls in wavelet applications.”
- Using wavelets to decompose the time–frequency effects of monetary policy (2008, unknown): “To perform the cross-wavelet analysis we could focus on the wavelet coherency, instead of the wavelet cross spectrum for three reasons: (1) there is some redundancy between the two measures, (2) the wavelet coherency has the advantage of being normalized by the power spectrum of the two time series,”