[ced]

The article had very few data points, that's why it worked with 3 dimensions. The deciding factor is how N (effective number of data points) compares with p (effective number of features).