Race as a Bundle of Sticks: Designs that Estimate Effects of Seemingly Immutable Characteristics

Although understanding the role of race, ethnicity, and identity is central to political science, methodological debates persist about whether it is possible to estimate the effect of something immutable. At the heart of the debate is an older theoretical question: Is race best understood under an essentialist or constructivist framework? In contrast to the “immutable characteristics” or essentialist approach, we argue that race should be operationalized as a “bundle of sticks” that can be disaggregated into elements. With elements of race, causal claims may be possible using two designs: (a) studies that measure the effect of exposure to a racial cue and (b) studies that exploit within-group variation to measure the effect of some manipulable element. These designs can reconcile scholarship on race and causation and offer a clear framework for future research.

Shall we count the living or the dead?

In the 1958 paper "Shall we count the living or the dead?", Mindel C. Sheps proposed a principled solution to the familiar problem of asymmetry of the relative risk. We provide causal models to clarify the scope and limitations of Sheps' line of reasoning, and show that her preferred variant of the relative risk will be stable between patient groups under certain biologically interpretable conditions. Such stability is useful when findings from an intervention study must be generalized to support clinical decisions in patients whose risk profile differs from the participants in the study. We show that Sheps' approach is consistent with a substantial body of psychological and philosophical research on how human reasoners carry causal information from one context to another, and that it can be implemented in practice using van der Laan et al's Switch Relative Risk, or equivalently, using Baker and Jackson's Generalized Relative Risk Reduction (GRRR).

Bayes-by Shower

Attention Check Items and Instructions in Online Surveys with Incentivized and Non-Incentivized Samples: Boon or Bane for Data Quality?

In this paper, we examine rates of careless responding and reactions to detection methods (i.e., attention check items and instructions) in an experimental sett

Are estimands being correctly used? A review of UK research protocols - Trials

Background The use of estimands in clinical trials was formalised with the adoption of the final International Conference on Harmonisation E9 Addendum on Estimands and Sensitivity Analysis in Clinical Trials (ICH E9(R1) Addendum) in November 2019. The declared objective of the ICH E9(R1) Addendum is to bring clarity and transparency to the research question of interest. For this to be achieved, the estimand must be described in accordance with the requirements of the ICH E9(R1) Addendum so that the target treatment effect is clear to all stakeholders. Previous reviews of publications and published protocols have found that few trials explicitly defined the primary estimand. To obtain a more complete picture of how the use of estimands has changed over time, whether trials are using estimands correctly (i.e. correctly defining the five attributes of an estimand), and which strategies are being used to handle intercurrent events, we obtained access to an extensive database of original research protocols (n = 29,212) submitted to the United Kingdom’s Health Research Authority, which oversees ethical review of clinical trials. Methods Protocols were eligible for review if they included the term ‘estimand’ and attempted to define at least one attribute of the primary estimand. For eligible protocols, we extracted information on trial characteristics such as whether the trial was randomized and the therapeutic area, as well as whether the estimand attributes used for the primary outcome were correctly defined, and which strategies were used to handle intercurrent events. Results We found that the number of protocols defining a primary estimand increased starkly with publication of the ICH E9(R1) Addendum (approximately 3 protocols/year pre-ICH E9(R1) Addendum vs. 18 protocols / year during the consultation period vs. 23 protocols in the year following the adoption of the ICH E9(R1) Addendum). However, the description of the primary estimand was suboptimal; many protocols failed to mention specific attributes (such as population or treatment conditions) in the estimand description, and many protocols incorrectly defined estimand attributes (e.g. by describing the estimand population based on their analysis population). Conclusions Although release of the ICH E9(R1) Addendum has dramatically increased the use of estimands in clinical trials, their reporting is suboptimal. There is still work to be done to ensure estimands reach their full potential in bringing clarity and focus to research questions.

Causal inference for N-of-1 trials

The aim of personalized medicine is to tailor treatment decisions to individuals' characteristics. N-of-1 trials are within-person crossover trials that hold the promise of targeting individual-specific effects. While the idea behind N-of-1 trials might seem simple, analyzing and interpreting N-of-1 trials is not straightforward. Here we ground N-of-1 trials in a formal causal inference framework and formalize intuitive claims from the N-of-1 trials literature. We focus on causal inference from a single N-of-1 trial and define a conditional average treatment effect (CATE) that represents a target in this setting, which we call the U-CATE. We discuss assumptions sufficient for identification and estimation of the U-CATE under different causal models where the treatment schedule is assigned at baseline. A simple mean difference is an unbiased, asymptotically normal estimator of the U-CATE in simple settings. We also consider settings where carryover effects, trends over time, time-varying common causes of the outcome, and outcome-outcome effects are present. In these more complex settings, we show that a time-varying g-formula identifies the U-CATE under explicit assumptions. Finally, we analyze data from N-of-1 trials about acne symptoms and show how different assumptions about the data generating process can lead to different analytical strategies.

Conceptualizing meaningful between-group difference in change over time: a demonstration of possible viewpoints

Is caviar a risk factor for being a millionaire?

Anders Huitfeldt argues that the answer depends on your definition of “risk factor” and calls for greater clarity in research The risk factor approach to epidemiology was introduced by the Framingham Heart Study investigators,1 2 who first alluded to the idea in 1951.3 The first use of the term “factor of risk” appeared in 1961,4 but it was not precisely defined. The resulting semantic confusion has hindered precise communication about study design and data analysis. To illustrate the problem, let us suppose that you want to study the causes and distribution of personal wealth. You have a secretive friend, and, among other questions, you are interested in knowing whether he is a millionaire. You are aware that there are some attributes, or risk factors, that are thought to be linked to being a millionaire. You decide to investigate. The first step is to choose your definition of risk factor. Clinical research can generally be divided into four broad objectives based on the intended use of the information obtained by the study: diagnosis, prognosis, treatment effects, and aetiology. Each of these research objectives is associated with a different definition. Table 1⇓ gives examples of how these four definitions of risk factor are used in the scientific literature and shows how each definition describes a different relation between the dependent variable and the independent variable. View this table: Table 1 Objectives of clinical research and associated definitions of risk factor A variable may qualify as a risk factor under more than one definition of the term. For example, cholesterol is believed to be a risk factor for heart disease under each of the four definitions. However, it is generally not plausible to assume that a variable that is a risk factor according to one definition will always be a risk factor under the …