The value of more information: A commentary on Stewart et al

• Published in: Social science & medicine (1982) 2010-12, Vol.71 (11), p.1910-1912
• Main Author: Lillard, Dean R.
• Format: Article
• Language: English
• Subjects: Bias; Cost-Benefit Analysis; Costs and benefits; Empirical Methods; Female; Females; Humans; Intervention; Low Income Groups; Mixed methods; Research Design; Revolutions; Sample selection; Selection Bias; Selection bias Sample selection Costs and benefits Mixed methods; Smoking; Smoking Cessation; Social Sciences; Values
• ISSN: 0277-9536; 1873-5347
• DOI: 10.1016/j.socscimed.2010.08.016

It has become common for social scientists to acknowledge & control for the bias & challenges to inferences that arise when one uses observational data from samples that selectively participate in the data generating process. Though social scientists have long been aware that selected samples produce bias, it was the seminal work of Heckman (1976,1979) that sparked a revolution in empirical methods to account for sample selection bias & that made the phrase common parlance. Such a discussion is critical because sample selection & selective attrition are so pervasive in observational data. In this short commentary I will discuss four shortcomings of "Impacts of a Support Intervention for Low-income Women Who Smoke" by Stewart et al. (2010). Because sample selection & sample attrition spawn a host of other problems, I focus mostly on the authors' failure to provide readers with the information they need to decide whether or not the analysis is fatally flawed. I will also briefly comment on a lack of discussion about some of the drawbacks of "holistic" research designs: the weakness of qualitative data: and costs & benefits of the studied smoking cessation intervention. While I comment on specific aspects of Stewart et al. (2010), the points apply quite broadly to studies that use similar methods. [Copyright Elsevier Ltd.]