© Institute of Mathematical Statistics, 2010 Assumptions of IV Methods for

By Vanessa Didelez, S. Meng, Nuala Sheehan

231 citations recorded by the source corpus.

Instrumental variable (IV) methods are becoming increas- ingly popular as they seem to offer the only viable way to overcome the problem of unobserved confounding in observational studies. However, some attention has to be paid to the details, as not all such methods target the same causal parameters and some rely on more restrictive parametric assumptions than others. We therefore discuss and contrast the most common IV approaches with relevance to typical applications in observational epidemiology. Further, we illustrate and compare the asymptotic bias of these IV estimators when underlying assumptions are violated in a numerical study. One of our conclusions is that all IV methods encounter problems in the presence of effect modification by unobserved confounders. Since this can never be ruled out for sure, we recommend that practical applications of IV estimators be accompa- nied routinely by a sensitivity analysis.

Subfields: Statistics and Probability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Statistics and Probability

Citations in the corpus

Causal inference by using invariant prediction: identification and confidence intervals (2015, neutral): “Instrumental variables can sometimes be used when the causal relationship of interest is confounded and no randomized experiments are available (Wright, 1928; Bowden and Turkington, 1990; Angrist et al., 1996; Didelez et al., 2010).”
Anchor regression: Heterogeneous data meet causality (2018, methodology): “Under IV assumptions (Didelez et al., 2010), lim →∞b = bcausal, that is one endpoint of anchor regression corresponds to the solution of Equation (2).”
Avoiding bias from weak instruments in Mendelian randomization studies. (2011, background): “In this paper, we consider continuous outcomes; with binary outcomes, weak instrument bias affects the causal estimate in a similar way, although it is not the only bias present [15].”
Causal Dantzig: Fast inference in linear structural equation models with hidden variables under additive interventions (2017, methodology): “We have mentioned already major diﬀerences to invariant causal prediction [Peters et al., 2016] and the loose relation to the vast literature on instrumental variable regression [Didelez et al., 2010] which will be detailed in Section 3.6.”
Multivariable Mendelian Randomization and Mediation. (2020, background): “Under additional assumptions described in detail elsewhere (Didelez et al. 2010) MR will give an estimate of the size of the causal effect of the exposure on the outcome.”
Detecting pleiotropy in Mendelian randomisation studies with summary data and a continuous outcome (2015, methodology): “The validity of the MR approach relies on three assumptions about the genetic variant used as an instrumental variable (IV): (1) the IV should be associated with the phenotype; (2) the IV should not be associated with any unobserved confounder of the association between phenotype and outcome; and (3”
A review of instrumental variable estimators for Mendelian randomization (2015, methodology): “With a continuous exposure, to identify a causal parameter, it is necessary to make some parametric assumptions, for example homogeneity and linearity in the causal exposure–outcome association with no effect modification.(9,12) The causal estimand is then the average change in the outcome for a uni”
Mendelian Randomization Analysis With Multiple Genetic Variants Using Summarized Data (2013, methodology): “For a binary outcome, such as disease, a log-linear or other appropriate regression model can be used in the regression of the outcome on the variant [Didelez et al., 2010].”