g

,

y

p

Figure 5.15 shows the relationship between the peak heights and

ues from the Gamma density model which was fitted to peak

or the smoothed peak spectrum shown in Figure 5.13. In this

model, those peaks with significantly small p values can be

ed as the true peaks or signals.

ver, each peak will not always be located in a single spectral

(the horizontal axis value in a spectrum). In other words, a peak

ficant peak may occupy an interval of the spectral analyser. This

at a peak may gradually move up from the bottom (the baseline)

ximum peak height and gradually move again from the maximum

ght down to the bottom. This can be seen in the smoothed peak

shown in Figure 5.13. Because of this, a merging process is taken

significant peak derived from the statistical learning model based

mma density model. This is to merge significant peaks to generate

spectra analyser value for each true peak. Only these merged

nalyser values will finally be treated as the signals. Figure 5.16(a)

example after significant peaks have been discovered using a

model fitted to peak heights. Figure 5.16(b) shows the result based

ttern shown in Figure 5.16(a) to discover the merged and unique

nt peaks.

(a) (b)

a) The significant peaks marked by the grey filled dots which were selected by

model fitted to the peak heights from a smoothed peak spectrum. (b) The merged

significant peaks, which are marked by the grey filled dots, which were

by merging the significant peaks shown in (a) of this figure.