nach oben

2010 | Buch

Kapitel lesen Erstes Kapitel lesen

Design of Observational Studies

verfasst von: Paul R. Rosenbaum

Verlag: Springer New York

Buchreihe : Springer Series in Statistics

Enthalten in: Springer Professional "Wirtschaft+Technik" , Springer Professional "Technik" , Springer Professional "Wirtschaft"

Einloggen, um Zugang zu erhalten

Über dieses Buch

An observational study is an empiric investigation of effects caused by treatments when randomized experimentation is unethical or infeasible. Observational studies are common in most fields that study the effects of treatments on people, including medicine, economics, epidemiology, education, psychology, political science and sociology. The quality and strength of evidence provided by an observational study is determined largely by its design. Design of Observational Studies is both an introduction to statistical inference in observational studies and a detailed discussion of the principles that guide the design of observational studies.

Design of Observational Studies is divided into four parts. Chapters 2, 3, and 5 of Part I cover concisely, in about one hundred pages, many of the ideas discussed in Rosenbaum’s Observational Studies (also published by Springer) but in a less technical fashion. Part II discusses the practical aspects of using propensity scores and other tools to create a matched comparison that balances many covariates. Part II includes a chapter on matching in R. In Part III, the concept of design sensitivity is used to appraise the relative ability of competing designs to distinguish treatment effects from biases due to unmeasured covariates. Part IV discusses planning the analysis of an observational study, with particular reference to Sir Ronald Fisher’s striking advice for observational studies, "make your theories elaborate."

The second edition of his book, Observational Studies, was published by Springer in 2002.

Inhaltsverzeichnis

Frontmatter

Beginnings

Frontmatter

Chapter 1. Dilemmas and Craftsmanship

Abstract

This introductory chapter mentions some of the issues that arise in observational studies and describes a few well designed studies. Section 1.7 outlines the book, describes its structure, and suggests alternative ways to read it.

Paul R. Rosenbaum

Chapter 2. Causal Inference in Randomized Experiments

Abstract

An observational study is an empiric investigation of treatment effects when random assignment to treatment or control is not feasible. Because observational studies are structured to resemble simple randomized experiments, an understanding of the role randomization plays in experiments is important as background. As a prelude to the discussion of observational studies in later chapters, the current chapter contains a brief review of the logic of causal inference in a randomized experiment. Only one simple case is discussed in detail, namely a randomized paired experiment in which subjects are paired before randomization and one subject in each pair is picked at random to receive treatment, the other receiving control. Although a foundation for later chapters, much of the material in this chapter is quite old, dating from Sir Ronald Fisher’s work in the 1920s and 1930s, and it is likely to be familiar from other contexts, such as a course in the design of experiments.

Paul R. Rosenbaum

Chapter 3. Two Simple Models for Observational Studies

Abstract

Observational studies differ from experiments in that randomization is not used to assign treatments. How were treatments assigned? This chapter introduces two simple models for treatment assignment in observational studies. The first model is useful but naïve: it says that people who look comparable are comparable. The second model speaks to a central concern in observational studies: people who look comparable in the observed data may not actually be comparable; they may differ in ways we did not observe.

Paul R. Rosenbaum

Chapter 4. Competing Theories Structure Design

Abstract

In a well designed experiment or observational study, competing theories make conflicting predictions. Several examples, some quite old, are used to illustrate. Also discussed are: the goals of replication, empirical studies of reasons for effects, and the importance of systemic knowledge in eliminating errors.

Paul R. Rosenbaum

Chapter 5. Opportunities, Devices, and Instruments

Abstract

What features of the design of an observational study affect its ability to distinguish a treatment effect from bias due to an unmeasured covariate u_ij? This topic, which is the focus of Part III of the book, is sketched in informal terms in the current chapter. An opportunity is an unusual setting in which there is less confounding with unobserved covariates than occurs in common settings. One opportunity may be the base on which one or more natural experiments are built. A device is information collected in an effort to disambiguate an association that might otherwise be thought to reflect either an effect or a bias. Typical devices include: multiple control groups, outcomes thought to be unaffected by the treatment, coherence among several outcomes, and varied doses of treatment. An instrument is a relatively haphazard nudge towards acceptance of treatment where the nudge itself can affect the outcome only if it prompts acceptance of the treatment. Although competing theories structure design, opportunities, devices, and instruments are ingredients from which designs are built.

Paul R. Rosenbaum

Chapter 6. Transparency

Abstract

Transparency means making evidence evident. An observational study that is not transparent may be overwhelming or intimidating, but it is unlikely to be convincing. Several aspects of transparency are briefly discussed.

Paul R. Rosenbaum

Matching

Frontmatter

Chapter 7. A Matched Observational Study

Abstract

As a prelude to several chapters describing the construction of a matched control group, the current chapter presents an example of a matched observational study as it might (and did) appear in a scientific journal. When reporting a matched observational study, the matching methods are described very briefly in the Methods section. In more detail, the Results section presents tables or figures showing that the matching has been effective in balancing certain observed covariates, so that treated and control groups are comparable with respect to these specific variables. The Results section then compares outcomes in treated and control groups. Because matching has arranged matters to compare ostensibly comparable groups, the comparison of outcomes is often both simpler in form and more detailed in content than it might be if separate adjustments were required for each aspect of each outcome. Treated and control groups that appear comparable in terms of a specific list of measured covariates – groups that are ostensibly comparable – may nonetheless differ in terms of covariates that were not measured. Though not discussed in the current chapter, the important issue of unmeasured covariates in this example is discussed in Part III.

Paul R. Rosenbaum

Chapter 8. Basic Tools of Multivariate Matching

Abstract

The basic tools of multivariate matching are introduced, including the propensity score, distance matrices, calipers imposed using a penalty function, optimal matching, matching with multiple controls and full matching. The tools are illustrated with a tiny example from genetic toxicology (n = 46), an example that is so small that one can keep track of individuals as they are matched using different techniques.

Paul R. Rosenbaum

Chapter 9. Various Practical Issues in Matching

Abstract

Having constructed a matched control group, one must check that it is satisfactory, in the sense of balancing the observed covariates. If some covariates are not balanced, then adjustments are made to bring them into balance. Three adjustments are almost exact matching, exact matching, and the use of small penalties. Exact matching has a special role in extremely large problems, where it can be used to accelerate computation. Matching when some covariates have missing values is discussed.

Paul R. Rosenbaum

Chapter 10. Fine Balance

Abstract

Fine balance means constraining a match to balance a nominal variable, without restricting who is matched to whom, when matching to minimize a distance between treated and control subjects. It may be applied to: (i) a nominal variable with many levels that is difficult to balance using propensity scores, (ii) a rare binary variable that is difficult to control using a distance, or (iii) the interaction of several nominal variables. The fine balance constraint and the distance can emphasize different covariates. When exact balance is unobtainable, fine balance can be used to obtain a specific pattern of imbalances.

Paul R. Rosenbaum

Chapter 11. Matching Without Groups

Abstract

Optimal matching without groups, or optimal nonbipartite matching, offers many additional options for matched designs in both observational studies and experiments. One starts with a square, symmetric distance matrix with one row and one column for each subject recording the distance between any two subjects. Then the subjects are divided into pairs to minimize the total distance within pairs. The method may be used to match with doses of treatment, or with multiple control groups, or as an aid to risk-set matching. An extended discussion of Card and Krueger’s study of the minimum wage is used to illustrate.

Paul R. Rosenbaum

Chapter 12. Risk-Set Matching

Abstract

When a treatment may be given at various times, it is important to form matched pairs or sets in which subjects are similar prior to treatment but avoid matching on events that were subsequent to treatment. This is done using risk-set matching, in which a newly treated subject at time t is matched to one or more controls who are not yet treated at time t based on covariate information describing subjects prior to time t.

Paul R. Rosenbaum

Chapter 13. Matching in R

Abstract

The statistical package R is used to construct several matched samples from one data set. The focus is on the mechanics of using R, not on the design of observational studies. The process is made tangible by describing it in detail, closely inspecting intermediate results; however, essentially, three steps are involved, (i) creating a distance matrix, (ii) adding a propensity score caliper to the distance matrix, and (iii) finding an optimal match. One appendix contains a short introduction to R. A second appendix contains short R functions used to create distance matrices used in matching.

Paul R. Rosenbaum

Design Sensitivity

Frontmatter

Chapter 14. The Power of a Sensitivity Analysis and Its Limit

Abstract

In an experiment, power and sample size calculations anticipate the outcome of a statistical test that will be performed when the experimental data are available for analysis. In parallel, in an observational study, the power of a sensitivity analysis anticipates the outcome of a sensitivity analysis that will be performed when the observational data are available for analysis. In both cases, it is imagined that the data will be generated by a particular model or distribution, and the outcome of the test or sensitivity analysis is anticipated for data from that model. Calculations of this sort guide many of the decisions made in designing a randomized clinical trial, and similar calculations may usefully guide the design of an observational study. In experiments, the power in large samples is used to judge the relative efficiency of competing statistical procedures. In parallel, the power in large samples of a sensitivity analysis is used to judge the ability of design features, such as those in Chapter 5, to distinguish treatment effects from bias due to unmeasured covariates. As the sample size increases, the limit of the power of a sensitivity analysis is a step function with a single step down from power 1 to power 0, where the step occurs at a value Γ̃ of Γ called the design sensitivity. The design sensitivity is a basic tool for comparing alternative designs for an observational study.

Paul R. Rosenbaum

Chapter 15. Heterogeneity and Causality

Abstract

Before R.A. Fisher introduced randomized experimentation, the literature on causal inference emphasized reduction of heterogeneity of experimental units. To what extent is heterogeneity relevant to causal claims in observational studies when random assignment of treatments is unethical or infeasible?

Paul R. Rosenbaum

Chapter 16. Uncommon but Dramatic Responses to Treatment

Abstract

Large effects in moderate to large studies are typically insensitive to small and moderate unobserved biases, but the concept of a ‘large effect’ is vague. What if most subjects are not much affected by treatment, but a small fraction, perhaps 10% or 20% of subjects, are strongly affected? On average, such an effect may be small, but not at all small for the affected fraction. Is such an effect insensitive to small and moderate unobserved biases?

Paul R. Rosenbaum

Chapter 17. Anticipated Patterns of Response

Abstract

Design sensitivity is used to quantify the effectiveness of devices discussed in Chapter 5. Several of those devices anticipate a particular pattern of results, perhaps coherence among several outcomes, or a dose-response relationship. To what extent do these considerations reduce sensitivity to unmeasured biases?

Paul R. Rosenbaum

Planning Analysis

Frontmatter

Chapter 18. After Matching, Before Analysis

Abstract

Three design tasks may usefully follow matching and precede planning of the analysis. Splitting the sample of I pairs into a small planning sample and a large analysis sample may aid in planning the analysis in a manner that increases the design sensitivity. If there will be analytic adjustments for some unmatched variables, it is prudent to check that the matched samples exhibit sufficient overlap on unmatched variables to permit analytic adjustments. Quantitative analysis of matched samples may usefully be combined with qualitative examination and narrative description of a few closely matched pairs.

Paul R. Rosenbaum

Chapter 19. Planning the Analysis

Abstract

“Make your theories elaborate” in observational studies, argued R.A. Fisher, so that the many predictions of such a theory may disambiguate the association between treatment and outcome. How should one plan the analysis of an observational study to check the predictions of an elaborate theory?

Paul R. Rosenbaum

Backmatter

Titel: Design of Observational Studies
verfasst von: Paul R. Rosenbaum
Verlag: Springer New York
Electronic ISBN: 978-1-4419-1213-8
Print ISBN: 978-1-4419-1212-1
DOI: https://doi.org/10.1007/978-1-4419-1213-8