scholarly journals Nonignorable Attrition in Pairwise Randomized Experiments

2021 ◽  
pp. 1-10
Author(s):  
Kentaro Fukumoto
Keyword(s):  
Causal Effect ◽  
Variance Estimator ◽  
Randomized Experiments

Abstract In pairwise randomized experiments, what if the outcomes of some units are missing? One solution is to delete missing units (the unitwise deletion estimator, UDE). If attrition is nonignorable, however, the UDE is biased. Instead, scholars might employ the pairwise deletion estimator (PDE), which deletes the pairmates of missing units as well. This study proves that the PDE can be biased but more efficient than the UDE and, surprisingly, the conventional variance estimator of the PDE is unbiased in a super-population. I also propose a new variance estimator for the UDE and argue that it is easier to interpret the PDE as a causal effect than the UDE. To conclude, I recommend the PDE rather than the UDE.

scholarly journals Treatment Effects on Ordinal Outcomes: Causal Estimands and Sharp Bounds

2018 ◽  
Vol 43 (5) ◽  
pp. 540-567 ◽  
Author(s):  
Jiannan Lu ◽  
Peng Ding ◽  
Tirthankar Dasgupta
Keyword(s):  
Causal Effect ◽  
Real Life ◽  
Causal Effects ◽  
Potential Outcomes ◽  
Randomized Experiments ◽  
Sharp Bounds ◽  
Average Causal Effect ◽  
Behavioral Studies ◽  
And Control ◽  
Potential Outcomes Framework

Assessing the causal effects of interventions on ordinal outcomes is an important objective of many educational and behavioral studies. Under the potential outcomes framework, we can define causal effects as comparisons between the potential outcomes under treatment and control. However, unfortunately, the average causal effect, often the parameter of interest, is difficult to interpret for ordinal outcomes. To address this challenge, we propose to use two causal parameters, which are defined as the probabilities that the treatment is beneficial and strictly beneficial for the experimental units. However, although well-defined for any outcomes and of particular interest for ordinal outcomes, the two aforementioned parameters depend on the association between the potential outcomes and are therefore not identifiable from the observed data without additional assumptions. Echoing recent advances in the econometrics and biostatistics literature, we present the sharp bounds of the aforementioned causal parameters for ordinal outcomes, under fixed marginal distributions of the potential outcomes. Because the causal estimands and their corresponding sharp bounds are based on the potential outcomes themselves, the proposed framework can be flexibly incorporated into any chosen models of the potential outcomes and is directly applicable to randomized experiments, unconfounded observational studies, and randomized experiments with noncompliance. We illustrate our methodology via numerical examples and three real-life applications related to educational and behavioral research.

scholarly journals Incorporating ethics and welfare into randomized experiments

2020 ◽  
Vol 118 (1) ◽  
pp. e2008740118
Author(s):  
Yusuke Narita
Keyword(s):  
Treatment Effect ◽  
Causal Effect ◽  
Well Being ◽  
Preference Elicitation ◽  
Controlled Trials ◽  
Randomized Experiments ◽  
Similar Treatment ◽  
Incentive Compatible ◽  
Randomized Controlled ◽  
Water Cleaning

Randomized controlled trials (RCTs) enroll hundreds of millions of subjects and involve many human lives. To improve subjects’ welfare, I propose a design of RCTs that I call Experiment-as-Market (EXAM). EXAM produces a welfare-maximizing allocation of treatment-assignment probabilities, is almost incentive-compatible for preference elicitation, and unbiasedly estimates any causal effect estimable with standard RCTs. I quantify these properties by applying EXAM to a water-cleaning experiment in Kenya. In this empirical setting, compared to standard RCTs, EXAM improves subjects’ predicted well-being while reaching similar treatment-effect estimates with similar precision.

Mediation: Causal Mechanisms in Business and Management

2021 ◽  
Author(s):  
Patrick J. Rosopa ◽  
Phoebe Xoxakos ◽  
Coleton King
Keyword(s):  
Statistical Power ◽  
Diversity Training ◽  
Causal Effect ◽  
Experimental Designs ◽  
Randomized Experiments ◽  
Modern Approach ◽  
Causal Mechanisms ◽  
Causal Mediation ◽  
Functional Forms ◽  
Traditional Approaches

Mediation refers to causation. Tests for mediation are common in business, management, and related fields. In the simplest mediation model, a researcher asserts that a treatment causes a mediator and that the mediator causes an outcome. For example, a practitioner might examine whether diversity training increases awareness of stereotypes, which, in turn, improves inclusive climate perceptions. Because mediation inferences are causal inferences, it is important to demonstrate that the cause actually precedes the effect, the cause and effect covary, and rival explanations for the causal effect can be ruled out. Although various experimental designs for testing mediation hypotheses are available, single randomized experiments and two randomized experiments provide the strongest evidence for inferring mediation compared with nonexperimental designs, where selection bias and a multitude of confounding variables can make causal interpretations difficult. In addition to experimental designs, traditional statistical approaches for testing mediation include causal steps, difference in coefficients, and product of coefficients. Of the traditional approaches, the causal steps method tends to have low statistical power; the product of coefficients method tends to provide adequate power. Bootstrapping can improve the performance of these tests for mediation. The general causal mediation framework offers a modern approach to testing for causal mechanisms. The general causal mediation framework is flexible. The treatment, mediator, and outcome can be categorical or continuous. The general framework not only incorporates experimental designs (e.g., single randomized experiments, two randomized experiments) but also allows for a variety of statistical models and complex functional forms.

scholarly journals A-14 Propensity Scores in Neuropsychological Research: Four Aspects of Digital Clock Drawing Distinguish Individuals with Non-Dementia Idiopathic Parkinson’s Disease from Matched Controls

2020 ◽  
Vol 35 (6) ◽  
pp. 787-787
Author(s):  
B Frank ◽  
C Dion ◽  
L Hizel ◽  
S Crowley ◽  
C Price
Keyword(s):  
Parkinson’S Disease ◽  
Parkinson's Disease ◽  
Bayes Factor ◽  
Propensity Scores ◽  
Nearest Neighbor ◽  
Causal Effect ◽  
Randomized Experiments ◽  
Digital Clock ◽  
Clock Drawing ◽  
Parsimonious Model

Abstract Objective In situations in which randomized experiments are impossible or unethical, propensity score matching offers a method to reduce bias on causal effect estimates (Thoemmes & Kim, 2011). In this study, we examined differences on the digital clock drawing test (dCDT; Souillard-Mandar et al., 2016) between individuals with idiopathic non-dementia Parkinson’s disease (PD) and matched controls. Method This study involved a retrospective analysis of two federally funded investigations (NSF-13-543; R01-NS082386). The sample included 261 participants (110 PD, 151 non-PD). Participants were matched according to demographic covariates, as well as measures of mood, comorbidity, and premorbid functioning. The PD group and matched controls were compared using logistic regression in a Bayesian framework, with projection predictive variable selection implemented to obtain a parsimonious model (Piironen, Paasiniemi, & Vehtari, 2018). All effects were standardized. Results Of 261 participants, 212 were matched using nearest neighbor matching (Figure 1). The final, parsimonious model included four variables from the dCDT: total strokes (command condition), total time (command condition), and area (command and copy conditions). While all effects were retained, positive to strong evidence was found for dCDT total time (βMedian = 0.91, βSD = 0.25, 95% CI [0.44, 1.42], Bayes factor [BF] = 97.80) and dCDT area (copy condition; βMedian = −0.52, βSD = 0.19, 95% CI [−0.90, −0.17], BF = 4.78). Conclusions Propensity scores can be employed in causal comparative studies to match control participants and reduce bias from nuisance covariates. Four aspects of dCDT performance were optimal in distinguishing individuals with PD from matched controls.

scholarly journals Bridging Finite and Super Population Causal Inference

2017 ◽  
Vol 5 (2) ◽  
Author(s):  
Peng Ding ◽  
Xinran Li ◽  
Luke W. Miratrix
Keyword(s):  
Causal Inference ◽  
Finite Population ◽  
Causal Effect ◽  
Causal Analysis ◽  
Potential Outcomes ◽  
Alternative Formulation ◽  
Randomized Experiments ◽  
Infinite Population ◽  
Variance Estimators ◽  
Average Causal Effect

AbstractThere are two general views in causal analysis of experimental data: the super population view that the units are an independent sample from some hypothetical infinite population, and the finite population view that the potential outcomes of the experimental units are fixed and the randomness comes solely from the treatment assignment. These two views differs conceptually and mathematically, resulting in different sampling variances of the usual difference-in-means estimator of the average causal effect. Practically, however, these two views result in identical variance estimators. By recalling a variance decomposition and exploiting a completeness-type argument, we establish a connection between these two views in completely randomized experiments. This alternative formulation could serve as a template for bridging finite and super population causal inference in other scenarios.

Estimating the Magnitude of the Relation Between Bullying, E-Bullying, and Suicidal Behaviors Among United States Youth, 2015

Crisis ◽  
2019 ◽  
Vol 40 (3) ◽  
pp. 157-165 ◽  
Author(s):  
Kevin S. Kuehn ◽  
Annelise Wagner ◽  
Jennifer Velloza
Keyword(s):  
Adolescent Suicide ◽  
Suicide Attempts ◽  
Causal Effect ◽  
Strong Association ◽  
Cross Sectional ◽  
Average Causal Effect ◽  
Effects Of Bullying ◽  
Pc Algorithm ◽  
Nationally Representative ◽  
Significant Covariates

Abstract. Background: Suicide is the second leading cause of death among US adolescents aged 12–19 years. Researchers would benefit from a better understanding of the direct effects of bullying and e-bullying on adolescent suicide to inform intervention work. Aims: To explore the direct and indirect effects of bullying and e-bullying on adolescent suicide attempts (SAs) and to estimate the magnitude of these effects controlling for significant covariates. Method: This study uses data from the 2015 Youth Risk Behavior Surveillance Survey (YRBS), a nationally representative sample of US high school youth. We quantified the association between bullying and the likelihood of SA, after adjusting for covariates (i.e., sexual orientation, obesity, sleep, etc.) identified with the PC algorithm. Results: Bullying and e-bullying were significantly associated with SA in logistic regression analyses. Bullying had an estimated average causal effect (ACE) of 2.46%, while e-bullying had an ACE of 4.16%. Limitations: Data are cross-sectional and temporal precedence is not known. Conclusion: These findings highlight the strong association between bullying, e-bullying, and SA.

The Effects of Questionnaire Length and Behavioral Consequences on Careless Responding

2020 ◽  
Vol 36 (2) ◽  
pp. 410-420 ◽  
Author(s):  
Anthony M. Gibson ◽  
Nathan A. Bowling
Keyword(s):  
Self Report ◽  
Randomized Experiments ◽  
High Quality ◽  
Buffering Effect ◽  
Careless Responding ◽  
Main Effect ◽  
Limited Support ◽  
Report Data ◽  
Self Report Data ◽  
Consistent Support

Abstract. The current paper reports the results of two randomized experiments designed to test the effects of questionnaire length on careless responding (CR). Both experiments also examined whether the presence of a behavioral consequence (i.e., a reward or a punishment) designed to encourage careful responding buffers the effects of questionnaire length on CR. Collectively, our two studies found (a) some support for the main effect of questionnaire length, (b) consistent support for the main effect of the consequence manipulations, and (c) very limited support for the buffering effect of the consequence manipulations. Because the advancement of many subfields of psychology rests on the availability of high-quality self-report data, further research should examine the causes and prevention of CR.

Optimal Experimental Design With Nesting of Persons in Organizations

2013 ◽  
Vol 221 (3) ◽  
pp. 145-159 ◽  
Author(s):  
Gerard J. P. van Breukelen
Keyword(s):  
Repeated Measures ◽  
Optimal Number ◽  
Health Centers ◽  
Randomized Experiments ◽  
Optimal Sample ◽  
Treatment Conditions ◽  
Sampling Cost ◽  
Nested Designs ◽  
Simple Equations ◽  
Number Of Individuals

This paper introduces optimal design of randomized experiments where individuals are nested within organizations, such as schools, health centers, or companies. The focus is on nested designs with two levels (organization, individual) and two treatment conditions (treated, control), with treatment assignment to organizations, or to individuals within organizations. For each type of assignment, a multilevel model is first presented for the analysis of a quantitative dependent variable or outcome. Simple equations are then given for the optimal sample size per level (number of organizations, number of individuals) as a function of the sampling cost and outcome variance at each level, with realistic examples. Next, it is explained how the equations can be applied if the dependent variable is dichotomous, or if there are covariates in the model, or if the effects of two treatment factors are studied in a factorial nested design, or if the dependent variable is repeatedly measured. Designs with three levels of nesting and the optimal number of repeated measures are briefly discussed, and the paper ends with a short discussion of robust design.

Sign in / Sign up

Export Citation Format

Share Document