PCA is one of a wide range of ordination methods available to ecologists. It is extremely useful as a hypothesis-generating tool, summarizing patterns in multivariate data, and reducing numbers of variables in analyses. It is best suited to data sets in which dependent variables are linearly related to each other, but may still provide useful results even if this condition is not fully met. Considerable control over the properties of the analysis can be exerted by use of data transformation and standardization, and it is important to be aware of the effects of these data manipulations on analysis interpretation. In general, raw data analyses will be influenced by variables with large values, whereas more extreme transformations or standardization will result in analyses in which variables with smaller values will exert stronger effects. PCA also has a range of other uses in multiple regression, and can provide summary variables that can be input into other analyses. It should not be blindly applied to data without consideration of data transformation and standardization, biplot scaling, and underlying relationships between variables. As with most ecological analyses, decisions made by the ecologist can alter the results and interpretation of numerical analysis; so it is important to ensure that correct decisions are made during the calculation of PCA.
See also: Ordination.
Was this article helpful?