How about pca and dimensionality reduction on binary data? Today it has come to my attention that doing pca on binary data might lose some of the trustworthiness I have done to associate with pca on continuous data. For reference see this discussion: pca on binary data - stack exchange