Understanding the Assumptions Underlying Instrumental Variable Analyses: a Brief Review of Falsification Strategies and Related Tools

Labrecque, Jeremy; Swanson, Sonja A.

doi:10.1007/s40471-018-0152-1

Cited by 132 publications

(139 citation statements)

References 52 publications

Supporting

Mentioning

136

Contrasting

Order By: Relevance

“…We use police agency dummy variables as exclusion restrictions in our heckprobit models for incarceration in jail or prison, which we might expect impact conviction but not sentencing (Bushway, Johnson, & Slocum, 2007). Since exclusion restrictions must be theoretically and substantively justified (Bushway et al., 2007; Hernán & Robins, 2006, 2017; Labrecque & Swanson, 2018), we used the literature and our knowledge of Miami's criminal justice process to select appropriate exclusion restrictions. We argue that the arresting agency helps to capture differences in the quality of evidence that might be related to conviction since some research has demonstrated differences in the evidentiary collection between law enforcement agencies (Roberts, 2015), but once convicted, the arresting agency should not influence incarceration decisions.…”

Section: Methodsmentioning

confidence: 99%

Institutionalizing inequality in the courts: Decomposing racial and ethnic disparities in detention, conviction, and sentencing*

Omori

Petersen

2020

Criminology

View full text Add to dashboard Cite

A significant body of literature has examined racial and ethnic inequalities in sentencing, focusing on how individual court actors make decisions, but fewer scholars have examined whether disparities are institutionalized through legal case factors. After finding racial and ethnic inequalities in pretrial detention, conviction, and incarceration based on 4 years of felony court data (N = 83,924) from Miami‐Dade County, we estimate nonlinear decomposition models to examine how much of the inequalities are explained by differences in criminal history, charging, and for conviction and incarceration, pretrial detention. Results suggest that inequality is greatest between White non‐Latinos and Black Latinos, followed by White non‐Latinos and Black non‐Latinos, ranging from 4 to more than 8 percentage points difference in the probability of pretrial detention, 7–13 points difference in conviction, 5–6 points in prison, and 4–10 points difference in jail. We find few differences between White non‐Latinos and White Latinos. Between half and three‐quarters of the inequality in pretrial detention, conviction, and prison sentences between White non‐Latino and Black people is explained through legal case factors. Our findings indicate that inequality is, in part, institutionalized through legal case factors, suggesting these factors are not “race neutral” but instead racialized and contribute to inequalities in court outcomes.

show abstract

Section: Methodsmentioning

confidence: 99%

Institutionalizing inequality in the courts: Decomposing racial and ethnic disparities in detention, conviction, and sentencing*

Omori

Petersen

2020

Criminology

View full text Add to dashboard Cite

show abstract

“…The monotonicity assumption: Z cannot increase X for some individuals and decrease it for others (e.g., Bollen, 2012 ; Labrecque & Swanson, 2018 ; Lousdal, 2018 ).…”

Section: Recommendations For Integrating Causality In a More Productimentioning

confidence: 99%

The Taboo Against Explicit Causal Inference in Nonexperimental Psychology

Grosz

Rohrer

Thoemmes

2020

Perspect Psychol Sci

250

238

View full text Add to dashboard Cite

Causal inference is a central goal of research. However, most psychologists refrain from explicitly addressing causal research questions and avoid drawing causal inference on the basis of nonexperimental evidence. We argue that this taboo against causal inference in nonexperimental psychology impairs study design and data analysis, holds back cumulative research, leads to a disconnect between original findings and how they are interpreted in subsequent work, and limits the relevance of nonexperimental psychology for policymaking. At the same time, the taboo does not prevent researchers from interpreting findings as causal effects—the inference is simply made implicitly, and assumptions remain unarticulated. Thus, we recommend that nonexperimental psychologists begin to talk openly about causal assumptions and causal effects. Only then can researchers take advantage of recent methodological advances in causal reasoning and analysis and develop a solid understanding of the underlying causal mechanisms that can inform future research, theory, and policymakers.

show abstract

“…IV1 is the only assumption which can be directly quantified (37); thus, we implement models (described below) to evaluate evidence for violations of these core assumptions. Specifically, pleiotropy, wherein a variant is associated with multiple phenotypes, may invalidate an IV if said pleiotropy constitutes an alternate causal pathway between the variant and the outcome (horizontal pleiotropy) (38).…”

Section: Selection Of Genetic Instrumental Variablesmentioning

confidence: 99%

Investigating the effect of glycaemic traits on the risk of psychiatric illness using Mendelian randomisation

Adams

Geaghan

Cairns

2020

Preprint

View full text Add to dashboard Cite

Data from observational studies have suggested an involvement of abnormal glycaemic regulation in the pathophysiology of psychiatric illness. This may be an attractive target for clinical intervention as glycaemia can be modulated by both lifestyle factors and pharmacological agents. However, observational studies are inherently confounded, and therefore causal relationships cannot be reliably established. We employed genetic variants rigorously associated with three glycaemic traits (fasting glucose, fasting insulin, and glycated haemoglobin) as instrumental variables in a two-sample Mendelian randomisation analysis to investigate the causal effect of these measures on the risk for eight psychiatric disorders. A significant protective effect of a unit increase in fasting insulin levels was observed for anorexia nervosa after the application of multiple testing correction (OR = 0.48 [95% CI: 0.33-0.71] -inverse-variance weighted estimate. The relationship between fasting insulin and anorexia nervosa was supported by a suite of sensitivity analyses, with no statistical evidence of instrument heterogeneity or horizontal pleiotropy. Further investigation is required to explore the relationship between insulin levels and anorexia.

show abstract

Understanding the Assumptions Underlying Instrumental Variable Analyses: a Brief Review of Falsification Strategies and Related Tools

Cited by 132 publications

References 52 publications

Institutionalizing inequality in the courts: Decomposing racial and ethnic disparities in detention, conviction, and sentencing*

Institutionalizing inequality in the courts: Decomposing racial and ethnic disparities in detention, conviction, and sentencing*

The Taboo Against Explicit Causal Inference in Nonexperimental Psychology

Investigating the effect of glycaemic traits on the risk of psychiatric illness using Mendelian randomisation

Contact Info

Product

Resources

About