How can statistical adjustment create a hypothetical analytic pseudo-population that differs from the actual experiences of the population?