# What's the name of this kind of circular visualization? (for high dimensional data, similar to PCA)

I'd like to ask what is the formal name of this kind of plot: http://www.wrcresearch.com/products/Slide6.PNG (in this case it's comparing different companies/brands based on several attributes) where the input is a multidimensional dataset and you want to see how these data compares in different dimensions, in a fashion similar to a principal components map, but with a circular layout.

And also if you know of any R package to produce this kind of visualization.

Thanks for all the answers! It seems to be PCA but I'm still not sure what adjustments to make so as to make it look circular like that. This is btw taken from the Excel plugin "BrandMap", and it looks very similar to a principal components maps of the same data: http://www.wrcresearch.com/products/Slide5.PNG

And the description from their webpage reads as "Concentric distance circles available for CGS or MCA maps." Although google doesn't suggest anything by those names.

• Looks like a biplot, a correlation biplot hence all arrows have unit length. And this could easily be produced from a PCA. Commented Jun 17, 2015 at 22:48
• @Gavin: if x and y coordinates of each arrow are given by correlation coefficients of the corresponding variable with the PC1/PC2 (I guess that's what you mean by "correlation biplot"), they do not necessarily have to be of unit length, do they? They cannot be longer, but they can be shorter. Here some seem to be a tiny bit shorter than the other, but overall they are remarkably close to the unit length; I guess this means that the dataset is pretty much two-dimensional. Commented Jun 17, 2015 at 23:09
• It seems that graphs which show canonical correlation analysis (CCA) results. Commented Jun 17, 2015 at 23:40
• @amoeba Gah, yes, sorry for that slip of the tongue. (By correlation biplot I mean a biplot drawn from results of PCA on the correlation matrix rather than the covariance matrix. Commented Jun 18, 2015 at 14:47
• @GavinSimpson, why not turn that into an official answer? Commented Oct 28, 2020 at 20:11