How can we reduce the number of variables? How can we use, say, two variables instead of six, without losing relevant information? One possibility is to exploit the correlations between pairs of variables: whenever this is high, both variables carry a similar amount of information and there may be no need of using both of them. Principal Component Analysis (PCA) is one of these techniques. Clearly, ordination is connected to ‘representation’ and it is aided by techniques that permit a reduction in the number of variables, with little loss in terms of descriptive ability. One main task of multivariate analysis is ordination, i.e. organising the observations so that similar subjects are near to each other, while dissimilar subjects are further away. Multivariate methods can help us to deal with multivariate datasets. For example, think about yield quality in genotype experiments or about the composition of weed flora in herbicide trials: in both cases, it is very important to consider several variables altogether, otherwise, we lose all information about the relationships among the observed variables and we lose the ability of producing a convincing summary of results. A main part of field experiments is multivariate, in the sense that several traits are measured in each experimental unit.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |