Risk factors, confounding and the illusion of statistical control

Research output: Contribution to journalArticle

Standard

Risk factors, confounding and the illusion of statistical control. / Christenfeld, NJS; Sloan, RP; Carroll, Douglas; Greenland, S.

In: Psychosomatic Medicine, Vol. 66, 01.01.2004, p. 868-875.

Research output: Contribution to journalArticle

Harvard

APA

Vancouver

Author

Christenfeld, NJS ; Sloan, RP ; Carroll, Douglas ; Greenland, S. / Risk factors, confounding and the illusion of statistical control. In: Psychosomatic Medicine. 2004 ; Vol. 66. pp. 868-875.

Bibtex

@article{5fdfe24361be4262b465d1250cc282df,
title = "Risk factors, confounding and the illusion of statistical control",
abstract = "ABSTRACT: When experimental designs are premature, impractical, or impossible, researchers must rely on statistical methods to adjust for potentially confounding effects. Such procedures, however, are quite fallible. We examine several errors that often follow the use of statistical adjustment. The first is inferring a factor is causal because it predicts an outcome even after {"}statistical control{"} for other factors. This inference is fallacious when (as usual) such control involves removing the linear contribution of imperfectly measured variables, or when some confounders remain unmeasured. The converse fallacy is inferring a factor is not causally important because its association with the outcome is attenuated or eliminated by the inclusion of covariates in the adjustment process. This attenuation may only reflect that the covariates treated as confounders are actually mediators (intermediates) and critical to the causal chain from the study factor to the study outcome. Other problems arise due to mismeasurement of the study factor or outcome, or because these study variables are only proxies for underlying constructs. Statistical adjustment serves a useful function, but it cannot transform observational studies into natural experiments, and involves far more subjective judgment than many users realize.",
keywords = "covariates, mediators, confounds, statistical control, risk factors",
author = "NJS Christenfeld and RP Sloan and Douglas Carroll and S Greenland",
year = "2004",
month = jan,
day = "1",
doi = "10.1097/01.psy.0000140008.70959.41",
language = "English",
volume = "66",
pages = "868--875",
journal = "Psychosomatic Medicine",
issn = "0033-3174",
publisher = "Lippincott Williams and Wilkins",

}

RIS

TY - JOUR

T1 - Risk factors, confounding and the illusion of statistical control

AU - Christenfeld, NJS

AU - Sloan, RP

AU - Carroll, Douglas

AU - Greenland, S

PY - 2004/1/1

Y1 - 2004/1/1

N2 - ABSTRACT: When experimental designs are premature, impractical, or impossible, researchers must rely on statistical methods to adjust for potentially confounding effects. Such procedures, however, are quite fallible. We examine several errors that often follow the use of statistical adjustment. The first is inferring a factor is causal because it predicts an outcome even after "statistical control" for other factors. This inference is fallacious when (as usual) such control involves removing the linear contribution of imperfectly measured variables, or when some confounders remain unmeasured. The converse fallacy is inferring a factor is not causally important because its association with the outcome is attenuated or eliminated by the inclusion of covariates in the adjustment process. This attenuation may only reflect that the covariates treated as confounders are actually mediators (intermediates) and critical to the causal chain from the study factor to the study outcome. Other problems arise due to mismeasurement of the study factor or outcome, or because these study variables are only proxies for underlying constructs. Statistical adjustment serves a useful function, but it cannot transform observational studies into natural experiments, and involves far more subjective judgment than many users realize.

AB - ABSTRACT: When experimental designs are premature, impractical, or impossible, researchers must rely on statistical methods to adjust for potentially confounding effects. Such procedures, however, are quite fallible. We examine several errors that often follow the use of statistical adjustment. The first is inferring a factor is causal because it predicts an outcome even after "statistical control" for other factors. This inference is fallacious when (as usual) such control involves removing the linear contribution of imperfectly measured variables, or when some confounders remain unmeasured. The converse fallacy is inferring a factor is not causally important because its association with the outcome is attenuated or eliminated by the inclusion of covariates in the adjustment process. This attenuation may only reflect that the covariates treated as confounders are actually mediators (intermediates) and critical to the causal chain from the study factor to the study outcome. Other problems arise due to mismeasurement of the study factor or outcome, or because these study variables are only proxies for underlying constructs. Statistical adjustment serves a useful function, but it cannot transform observational studies into natural experiments, and involves far more subjective judgment than many users realize.

KW - covariates

KW - mediators

KW - confounds

KW - statistical control

KW - risk factors

UR - http://www.scopus.com/inward/record.url?scp=9244223046&partnerID=8YFLogxK

U2 - 10.1097/01.psy.0000140008.70959.41

DO - 10.1097/01.psy.0000140008.70959.41

M3 - Article

C2 - 15564351

VL - 66

SP - 868

EP - 875

JO - Psychosomatic Medicine

JF - Psychosomatic Medicine

SN - 0033-3174

ER -