Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

A Source of False Findings in Published Research Studies

A Source of False Findings in Published Research Studies Opinion EDITORIAL Adjusting for Covariates Helena Chmura Kraemer, PhD Concern about erroneous conclusions of many published re- The linear model used for covariate adjusting (eg, analy- search findings has led to the conclusion that most published sis of covariance [ANCOVA]) assumes, for all possible values 1,2 research findings are wrong. What can be done about that? of the covariates, that covariate ES is equal to typical ES; that In what follows, I will focus on one common source of false is, that there is no interaction between the covariates and the findings: adjusting for covariates. treatment effect. If this assumption is violated, then the in- Here, adjusting means allowing variables to vary as they teractions that exist in the population (but are not included will but then using a mathematical model to assess their in- in the model) can bias the statistical tests and estimation of fluence on the outcome. In contrast, to control means manipu- the treatment ES. Furthermore, not finding statistically sig- lation of variables by the researcher for a particular purpose nificant interactions in the sample does not prove the null hy- (eg, in experimental design). Unfortunately, the terms adjust pothesis that they do not exist http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png JAMA Psychiatry American Medical Association

A Source of False Findings in Published Research Studies

JAMA Psychiatry , Volume 72 (10) – Oct 1, 2015

Loading next page...
 
/lp/american-medical-association/a-source-of-false-findings-in-published-research-studies-Hcv050V1AW
Publisher
American Medical Association
Copyright
Copyright 2015 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.
ISSN
2168-622X
eISSN
2168-6238
DOI
10.1001/jamapsychiatry.2015.1178
pmid
26244634
Publisher site
See Article on Publisher Site

Abstract

Opinion EDITORIAL Adjusting for Covariates Helena Chmura Kraemer, PhD Concern about erroneous conclusions of many published re- The linear model used for covariate adjusting (eg, analy- search findings has led to the conclusion that most published sis of covariance [ANCOVA]) assumes, for all possible values 1,2 research findings are wrong. What can be done about that? of the covariates, that covariate ES is equal to typical ES; that In what follows, I will focus on one common source of false is, that there is no interaction between the covariates and the findings: adjusting for covariates. treatment effect. If this assumption is violated, then the in- Here, adjusting means allowing variables to vary as they teractions that exist in the population (but are not included will but then using a mathematical model to assess their in- in the model) can bias the statistical tests and estimation of fluence on the outcome. In contrast, to control means manipu- the treatment ES. Furthermore, not finding statistically sig- lation of variables by the researcher for a particular purpose nificant interactions in the sample does not prove the null hy- (eg, in experimental design). Unfortunately, the terms adjust pothesis that they do not exist

Journal

JAMA PsychiatryAmerican Medical Association

Published: Oct 1, 2015

References