I discuss my research on vaccination planning. Social interactions in treatment response make infectious disease a core concern of public health policy. Spread of infection creates a negative external effect. Preventive administration of vaccines may reduce disease transmission. In a decentralized healthcare system, infected and at-risk persons may not adequately recognize the social implications of their actions. Hence, there may be a rationale for government to seek to influence treatment of infectious diseases. Policies range from quarantines of infected persons to mandatory vaccination to subsidization of vaccines and drugs. I focus on a prevalent difficulty, being scarcity of evidence about how interventions affect illness. Randomized trials, which have been central to evaluation of treatments for noninfectious diseases, are less informative about treatment of infectious diseases. I develop minimax regret policy based on credible assumptions. I first consider a simple representative-agent setting in which members of a large population share identical cost of vaccination, cost of illness, probability of vaccine effectiveness, and probability of illness when unvaccinated or unsuccessfully vaccinated. I then generalize to vaccination of a heterogeneous population.

Section 7.1 examines the idealized setting where optimal utilitarian planning is feasible. I quantify the value of covariate information in improving achievable social welfare. Conditioning treatment choice on more refined covariate information cannot lower social welfare in this setting. It increases welfare if covariate refinement has predictive power that affects treatment choice. I also discuss nonutilitarian arguments to disregard certain covariate information when making clinical and criminal justice decisions.

The remainder of the chapter considers planning when uncertainty about treatment response makes optimization infeasible. Section 7.2 distinguishes settings where the planning problem does or does not decompose into a set of separable covariate-specific problems. The former situations are easier to study than the latter. Section 7.3 considers the common medical problem of choice between surveillance and aggressive treatment of patients, with partial knowledge of personalized risk of illness. Section 7.4 extends the analysis to sequential choice of whether to acquire costly covariate information as a prelude to treatment choice. I focus on medical choice to perform a diagnostic test before making the treatment decision.

Section 10.1 calls for work that strengthens the foundations for planning under uncertainty in three ways: communicating uncertainty, specifying planning problems, and enhancing the tractability of decision criteria. Among the many substantive problems that warrant study, Section 10.2 cites the immediate need for improved pandemic planning, whose salience society should appreciate following the recent global experience with COVID-19. Section 10.3 concludes with a personal perspective on an existential societal decision, making public a commentary that I wrote forty years ago but have not circulated until now.

I mainly discuss revealed preference analysis, which assumes that choice results from maximization of personal welfare. Section 4.1 discusses revealed preference analysis in some generality. Section 4.2 focuses on identification of income-leisure preferences for evaluation of income tax policy. Whereas Sections 4.1 and 4.2 concern behavior in deterministic settings, Section 4.3 considers identification when it is assumed that individuals cope with uncertainty by maximizing expected utility. Going beyond revealed preference analysis, Section 4.4 discusses identification using subjective data in place of or in addition to observations of actual choices.

I first describe how the physical-science and economics literatures have sought to cope with uncertainty about the correct climate model and discount rate, respectively. I next formalize MMR policy choice in an abstract manner. I then present the computational model studied in my research and summarize the main findings.

I first review and critique the prevailing use of hypothesis tests to compare treatments. I then describe my application of statistical decision theory. I compare Bayes, maximin, and minimax regret decisions. I consider choice of sample size in randomized trials from the minimax regret perspective.

Section 3.1 presents key findings on partial identification of mean treatment response with observational data, with accompanying illustrations. Section 3.2 considers identification problems that arise with data from randomized trials, focusing on an important practical issue that has escaped notice until recently. Section 3.3 discusses the difficult problem of identification of treatment response with social interactions. Section 3.4 revisits the subject of meta-analysis discussed in Chapter 2, now offering an alternative to the manner in which meta-analysis has been performed to date.

This chapter introduces some nonlinear time series models of widespread use in economics and finance. Specifically, we consider structural breaks, GARCH models, and copula models.

This chapter gives a more comprehensive treatment of nonparametric methods for estimating density functions and dynamic regression models. We also consider the emerging material on the case where there are many explanatory variables and how selection methods can be used to apply estimation and inference techniques to this case.