Question

What is selection bias in sampling? How can it be controlled? How does selection bias affect the potential use of evidence yielded by research that has a sample with selection bias? Please include a reference. Thank you​

Answer 1

If a researcher fails to achieve proper randomization while sample selection, a bias can occur. Improper randomization leads to the collection of samples that are not a true representative of the population intended to be analyzed. It is otherwise known as a selection effect. The results of the study may not be accurate unless the selection bias is controlled. Enrolling of the target population as such in the study is not possible so that the researcher should obtain fair samples with a minimal bias to obtain a true measure of association. I other words the sample must be the true representation of the population.

Types of sampling bias:

• Attrition bias (loss-to-follow-up)
• Healthy entrant effect
• Response bias
• Survivorship bias
• Confounding bias

Methods to control selection Bias:

• Adjust for Selection Bias: Include all variables associated with selection in the analysis to reduce selection bias in a similar way to confounding [

• Bias Breaking: Use of odds ratio in case-control studies.

• Imputation: Replacement of missing values with a reasonable estimate using the collected data.

• Use of population data in a case control study instead of control data.

• Post-Stratification:   Use of stratified sampling or frequency matching.

• Prediction of the Bias: Obtain information from non-participants to obtain information about the amount of bias present.

• Propensity Score: Use as a covariate during analysis or match cases and controls
• Sensitivity Analysis: to assess the direction and magnitude of the bias.

• Stratification: Helps to obtain unbiased estimates within strata or group.

• Weighting: Like inverse probability weighting to represent non-participants.

In order to attain validity of the result, the researcher must design studies to minimize the chances of bias. During the analysis, the researcher must identify whether bias may have been introduced at any stage. Appropriate statistical methods could be used to curtail the effect of bias.

References:

1. Claire Keeble1, Graham Richard Law1, Paul D. Baxter, Stuart Barber. Choosing a Method to Reduce Selection Bias: A Tool for Researchers, Open Journal of Epidemiology, 2015, 5, 155-162

Published Online August 2015 in SciRes. http://www.scirp.org/journal/ojepi

http://dx.doi.org/10.4236/ojepi.2015.53020

2. Dr. Carl M. Shy, Epidemiology 160/600 Introduction to Epidemiology for Public Health course lectures, 1994-2001, The University of North Carolina at Chapel Hill, Department of Epidemiology