18.08.2013 Views

THESIS

THESIS

THESIS

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

1.2.2 Cluster analysis<br />

I. Ratio scale traits (quantitative traits)<br />

The 69 accessions belonging to Solanum spp. was analysed based on<br />

the 13 traits selected using the PROC CLUSTER statement. This procedure searched<br />

throughout the data for accessions that were similar enough to be considered as part of<br />

a common cluster (Cababasay, 1996).<br />

Using the standardized data, the Unweighted Pair Group of<br />

Matheretic Average (UPGMA) was chosen to stratify the 69 accessions. This could be<br />

attributed to the fact that average linkage method (UPGMA) can be used with any<br />

resemblance coefficients compared with Ward’s method. Moreover, the similarity<br />

between pairs of clusters can be determined in a manner less extreme compare with<br />

single or complete linkage (Romesburg, 1984). This was done by taking the average<br />

distance between all pairwise combinations of observations one in each cluster (SAS,<br />

1999). This method was also recommended by Cababasay (1991) due to its best overall<br />

performance.<br />

The PROC CLUSTER statement provided useful statistics (pseudo<br />

F, pseudo t 2 and cubic clustering criteria) that can estimate the number of clusters that<br />

will be formed out of a given data. They refer to the values that are measures of fit for<br />

the analysis of each specified number of clusters (Callantes, 2003). These statistics<br />

were given using the PSEUDO and CCC options, respectively, which are one of the<br />

options used to control the printing of the cluster history (SAS, 1999).<br />

With pseudo F statistic, Figure 3 reveals that the possible numbers<br />

of clusters used to stratify the 69 accessions of Solanum spp. were 2 and 10. With the<br />

pseudo t 2 statistic, however, Figure 4 shows that the accessions could be grouped in to<br />

1 and 6 clusters. Figure 5, on the other hand, gave an estimate of 7 clusters using the<br />

cubic clustering criterion. In this case, there was no consensus that could be found<br />

among these criteria.<br />

48

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!