What is the distribution of numerical feature values across the samples? What is the distribution of categorical features?