Numerical evaluation of cytologic data. IX. Search for data structure by principal components transformation.
Bartels Ph · PubMed · 1981
Principal components transformation may be used to explore the structure of a p-dimensional data set. It is difficult to detect inhomogeneities in a data set of multivariate variables by mere visual inspection of the numerical data. Plotting each variable's distribution is often either impractical, due to the number of variables involved, or might fail to reveal the presence of subpopulations due to high correlations. A practical example is given in which principal components transformation revealed the presence of subpopulations in a four-dimensional data set.