Only the (E(MV,LT,ST)1,db7 +, E(MV,LT,ST)1,db7 −) correlation was less than 0.9 ( Figure 6); in the other cases it was close to 1. It was shown that only the first two energies calculated for db7 wavelets yielded suitable results, because for higher scaling parameters they SGI-1776 were correlated with wavelet energies calculated from mexh. It was decided to

add three additional parameters, besides the energies for db7, defined as: equation(18) Ei,db7±=Ei,db7++Ei,db7−2fori=1,2E1,|db7|=|E1,db7+−E1,db7−| for every deviation type MV, LT and ST. For the fractal dimension, the quality of the results obtained using semivariograms, spectral and wavelet analyses was insufficient. Box size counts were found to be the most efficient methods. The application of a median filter to bathymetric profile segments was also a good way of finding diverse forms on the example SB431542 cost profile (Figure 7). The above analyses demonstrate that to describe the diverse morphology of Brepollen the following parameters have to be taken into account: M0, M1, M2, M3, γ, E1, mexh, E2, mexh, E3,

mexh, E4, mexh, E5, mexh, E6, mexh, E7, mexh, E1, db7 ±, E2, db7 ±, E1, |db7|, Dbox, MF1, MF2, MF3, MF4, MF5, MF6. As these parameters could still be independent, the input parameters were reduced by Principal Component Analysis (PCA). Before embarking on PCA, the distributions of the values of each parameter were analysed. Two types of calculated values were identified: (i) with data where quantity is encompassed within one order of magnitude (γ, Dbox, MF1, MF2, MF3, MF4, MF5, MF6) and (ii) with data whose values range over several orders of magnitude (M0, M1, M2, M3, E1, mexh, E2, mexh, E3, mexh, E4, mexh, E5, mexh, E6, mexh, E7, mexh, E1, db7 ±, E2, db7 ±, E1, |db7|). For the second case the common logarithm was determined. The next step included data normalisation: equation(19) xm=x−xsrσx, where xm – new parameter value, x – its determined value, xsr – mean value of determined parameters, σx– standard deviation

of determined parameters. After such parameter transformation, the mean of each one will be equal to zero and the standard deviation equal to one. Analysis of the variance of Principal Megestrol Acetate Components (PCs) (Figure 8) showed their diminishing influence on the overall value. For the independent analysis of every deviation, the first ten PCs are sufficient for cluster analysis. Together, these correspond to more than 98% of the cumulative variance. In the analysis of deviation MV, this value was exceeded by the first nine PCs, but despite this, it was decided to use the same number as in the other two cases. When all the parameters were included, 98% of the cumulative variance was exceeded for the first 16 PCs, and this number of parameters was utilised in the cluster analysis.