what's special about smartpca vs regular pca

Question

[Zhenyu Zhang]

I know people use smartpca to check outliers and also save eigenvectors for population structures, and it seems to be a standard. My question is that what's so smart about smartpca? Is the result or method significantly different from any PCA packages, for example, pccomp in R? or does smartpca use some specially modeling methods where we can not find in other packages?

[Tom K]

I think smartpca is just meant to be faster than prcomp. There are many implementations of PCA and some are needed to run larger matrices that require too much compute-time or memory to run with regular approaches. They may give slightly different results due to approximations (which are usually acceptable provided you detail you methods when publishing).