Figure 3

Cartoon depicting the process used to identifypeak-bags. Each panel shows the same example wavelet analysis output (bottom), with peaks coloured corresponding to the originating sub-spectra, and a hypothetical p-value scale space (top), where white regions are termed significant. (a) Analysis over multiple window sizes. Window size 5 at m/z = 10 is significant, but has multiple contributions from the same sub-spectrum (i.e., the black peaks). Therefore, a window of size 3 is used for the new peak-bag, since it is the largest significant window size with at most one contribution from each sub-spectra. Scale space elements that, when used as a future window-center, reference peak positions lying in the current window, are greyed out and the process restarts. (b) Although the combined peaks are significant, there are again multiple contributions from one sub-spectrum at window size 3. Position 5 at window size 1 is not significant. (c) Peak-bag at slightly higher scale. Larger windows are greyed out, so a new peak-bag is selected with window size 5. (d) No significant elements remaining.