This article presents a scenario for the future of research access to federally collected microdata. Many researchers find access to government databases increasingly desirable. The databases themselves are more comprehensive, of better quality and-with improved database management techniques--better structured. Advances in computer communications enable remote access to these databases. Substantial gains in the performance/cost ratio of computers permit more sophisticated analyses--including ones based on statistical graphics, identification of extreme or influential values, record linkage and Bayesian regression methods. At the same time, the individuals and institutions that provide the data residing on government databases--as well as the agencies who sponsor the collection of such information--are becoming increasingly aware that the same technologies that extend analytical capabilities also furnish tools that threaten the confidentiality of data records. As the broker between the data provider and the data user, government agencies are under increased pressure to implement policies that both increase data access and ensure confidentiality. In response to these cross-pressures, agencies will more actively pursue statistical, administrative and legal approaches to responsible data dissemination. Recent developments in these approaches are discussed as they relate to improvements in database techniques, computer and analytical methodologies and legal and administrative arrangements for access to and protection of federal statistics.

In many real-life problems, avoiding trouble can be at least as important as achieving optimality and is a necessary precondition anyway. Some guidelines are suggested for avoiding pitfalls throughout a statistical investigation. A range of real-life examples is presented to show how difficulties can arise in practice and how they may be overcome. These examples are unlike the idealized and sanitized illustrations that typically appear in the literature.

Population models in the birth-and-death style tradition have the unpleasant (and usually not advertised) implication that individuals do not age: It follows from the Markov properties of the whole population that life spans must be exponentially distributed and reproduction occur as splitting or in a Poisson process. This can be remedied only in parts (and at a high esthetical cost) by assuming more complicated Markovian properties in real time, like the age and parity dependent models of demography. Instead, if there is a sensible Markov structure in population growth, it resides in the pedigree, daughters inheriting genotypes from their mothers and being independent of their ancestors, given these types. This idea is used to define general branching processes and to analyze their properties: extinction, growth and asymptotic composition. The results are used to interpret the hypothesis of a molecular clock of mutations in biological evolution.

De Moivre gave a simple closed form expression for the mean absolute deviation of the binomial distribution. Later authors showed that similar closed form expressions hold for many of the other classical families. We review the history of these identities and extend them to obtain summation formulas for the expectations of all polynomials orthogonal to the constants.