The reliability of methods for predicting missing figures in migration tables (Q791270)

We try to supplement existing methods by providing prediction intervals for missing figures. Our aim is that about (1-\(\alpha)\) 100\% of these intervals contain the true outcome of the corresponding figure. This requires complete specification of underlying probabilistic assumptions. In section 2 an initial guess \(P_ 0\) is extended to a convenient family. This enables us to provide prediction intervals. However, our assumptions are so restrictive that the reliability of the proposed methods is questionable. This is illustrated in section 3 by means of a practical example where the predicted figures were actually known.