| Author | Message | ||
     Michel Coene (Michel) |
I often observe that in a 2D scoresplot (PC1 vs PC2), part of my samples lie on a perfect circle segment or on a hyperbole. This is usually only visible if you don't plot with equal scales on X and Y. Anybody any idea what this means? | ||
| hlmark |
Michel - this is discussed on pages 179-180 of Rich Kramer's "Chemometric Techniques for Quantitative Analysis". Marcel Dekker (1998) In his book, Kramer uses the term "normalization" to describe the operation leading to this display, which he puts in the context of deliberately trying to obtain this result, which can be useful under some circumstances. If it's happenning to you without your intendiing it to, and you consider it an undesirable result, then you have to look for the cause; the possible causes can be seen in Richard's discussion: there's not enough variabilty in the original data, one or more pretreatments applied to the data are removing useful information, too many factors (it's not clear whether your PC1 and PC2 are in fact tha first two PC factors, or just any two factors), etc. Howard |