why might s be more commonly used than s squared as a measure of variability?