Comment by jmount

Small proviso: this is only true for a reasonable number of categories (or you run into repeated experiment problems).