Hi all,
I think I may be confused by different people/programs using the word
rotation differently.
Does prcomp not perform rotations by default?
If I understand it correctly retx=TRUE returns ordinated data, that I can
plot for individual samples (prcomp()$x: which is the scaled and centered
(rotated?) data multiplied by loadings).
What does it mean that the data is rotated from the "?prcomp"
description?
Is this referring to the data matrix orientation (i.e. looking at
differences among samples (columns) based on variables (rows) vs.
differences among variables (columns) based on samples(rows))?
Thank you,
Colin Wahl
Graduate student,
Western Washington University
code & background:
I am looking at the ordination of abiotic stream variables between different
sampling locations.
abiot.pca=prcomp(all24[, c(10, 13:18)], retx=TRUE, center=TRUE, scaleTRUE)
summary(abiot.pca)
Importance of components:
PC1 PC2 PC3 PC4 PC5 PC6 PC7
Standard deviation 1.5925 1.0814 1.0697 0.9536 0.76624 0.68444 0.43037
Proportion of Variance 0.3623 0.1671 0.1635 0.1299 0.08387 0.06692 0.02646
Cumulative Proportion 0.3623 0.5294 0.6928 0.8227 0.90662 0.97354 1.00000
loadings[,1:3]
PC1 PC2 PC3
avg.dmax -0.1879223 0.55792480 -0.04962935
scond.med -0.4327223 0.04779998 -0.43369560
docon.med 0.1976094 -0.30384127 0.67222550
cb.per 0.4134302 -0.17550281 -0.40171318
gc.per 0.4136933 -0.26997129 -0.39960398
gf.per 0.3419349 0.63917223 0.16174661
fine.per -0.5285840 -0.28616200 0.10168155
[[alternative HTML version deleted]]
