With fading support, use measures of variability for numerical data from random samples to draw X informal comparative inferences about two populations; X out of X times, with X% accuracy.