Wavelet analysis works most convincingly on complicated data sets. Then, the mapping of the data into the plane of time and duration displays in various ways the richness and complexity of the data. These wavelet maps promptly represent an overwhelming amount of information, that must be reduced in order to be useful.
The reduction can take two basic different forms, and countless variants. In the sift-and-discard approach, features of interest are enhanced and the rest is removed from the wavelet maps: applications such as denoising or event identification are representative. In the statistical approach, duration-dependent distributions and moments are calculated. Conditional statistics combine the two approaches. The user will probably need to devise his or her own algorithms to match the goals of each application. This section is merely illustrative of the contents of Pandora's box.