6.7 Takeways

• evenly-spaced data are handled with the CWT algorithm of Torrence & Compo (1998), which is very fast; unevenly-spaced data are handled through the WWZ algorithm of Foster (1996), which is slow. The temptation to interpolate is strong, but it has serious downsides (INTERPOLATION = MAKING UP DATA!). WWZ should be preferred, though CWT+interpolation can be used at exploratory stages.

• Wavelet Transform Coherency may be thought of as a (squared) correlation coefficient in time-scale space: 0 if the series do not covary, 1 if they vary in perfect unison at that time and scale.

• Properly interpreting the results of wavelet and wavelet coherence analyses requires knowing something about the methods.