A few months ago there was a kerfuffel about selection bias in the statistical analysis of fmri data points for brain scans. Now here’s another complication: selection bias in populating the studies. People afraid of getting blood drawn and going into an frmi don’t sign up for brain scan studies. The authors suggest using people who already have had some kind of imaging; surely they won’t mind another. The only trouble is that these folks may not be normals; even musculo skeletal harm can skew cognitive results. Here’s the soon-to-be-published abstract from SSRN:
Roe, Brian E., Haab, Timothy C., Beversdorf, David Q., Gu, Howard H. and Tilley, Michael R.,Risk-Attitude Selection Bias in Subject Pools for Experiments Involving Neuroimaging and Blood Samples (March 4, 2009). Journal of Economic Psychology, Forthcoming. Available at SSRN:http://ssrn.com/abstract=1353104:
Abstract:
Techniques such as neuroimaging and molecular genetics are increasingly used to investigate economic theory, decision making behavior and personality traits related to economic behavior (e.g., risk attitudes, reward dependence). . . .. We find recruits with more conservative risk attitudes in two of four measured dimensions are less likely to agree to participate in the study due to these biomedical requirements [a blood sample and an mri scan], suggesting that recruitment among student volunteer populations for fMRI studies and for genetics studies requiring blood as genetic source material may induce a sample selection bias in the domain of risk attitudes. We find that limiting recruitment to individuals who have previously undergone certain types of medical interventions (MRI, computed tomography or surgery) eliminates the sample selection bias in the case of fMRI research and attenuates the bias in the case of genetics research. Furthermore, relying upon buccal cells rather than blood for genetic source material may attenuate sample selection bias. Buccal cell samples can be collected via less invasive oral techniques and have been shown to provide genotyping results that are comparable to blood samples.
Comments