An investigation on principal component, discriminant and cluster analyses and their application on ecological data

There may be many interrelated variables affecting a result reached by any study. Most of the problems challenging the contemporary researchers studying on biological researches are related to whether there is any relationship between two or more variables. Therefore, in order to achieve a reliable and accurate study, it is essential to determine the relations between two or more variable clutters, and that makes the analysis techniques, provided this becomes more important. The aim of this study is to examine Principle Component Analysis, discriminant analysis and cluster analysis in detail, all of which are among multi-variable statistical analyses. The statistical methods were applied to some ecological data, through which it was aimed to understand how the results must be evaluated and thus to reach a better understanding of the theoretical and practical view of the subject. Principle Component Analysis were applied to 68 species caught during a one-year catching season, and the species composing the majority of the biomass were found to be peregrineshrimp, lizardfish, red mullet, goldband goatfish and crab. Similar results were reached via cluster analysis at the 90% similarity level. Discriminant analysis was used to understand whether the brown comber caught from Mersin Babadillimani and Iskenderun Bay are the same stocks. BoxM test was applied to the morphometric features of these fish to find out whether linear or quadratic discriminant analysis should be preferred. As a result, linear discriminant analysis was determined to apply only to the data of February. Using discriminant analysis, two regions were seen to include two different stocks, and the ratio of the caught fish belonged to their own stock were 98.3% in January, 80.4% in February, 100% in March, 94% in April, 98.6% in May, 92.3% in June, 92.9% in July, 82.5% in August, 89.7% in September, 98.8% in November and 92.2% in December. Key Words: Principal component, discriminant and cluster analyses