Overview
Selection Bias occurs when the sample that is drawn to represent a population does not match the population in question. As a general rule, the sampled population will never match the population at large due to random factors, however, selection bias is usually concerned with systemic factors which tilt every sampled population away from the true population.
The standard example of selection bias occurs when individuals contacted for a survey either willingly take a survey, or refuse to do so. As a matter of course, the researcher will assume (hope) that the willingness to take a survey is uncorrelated with the questions on the survey. If there is correlation between the willingness to take a survey and the questions on the survey, the results of the survey will be biased.
Correcting Selection Bias
- Weighted Samples - if a selection bias is known to exist which causes certain demographics to be overweighted or underweighted in a sample, one method of correcting the bias is to weight the samples differently from the different groups when reporting the statistics of the chosen sample. The over/under weighting is typically determined by looking at previous surveys and determining what the selection bias in those surveys were (if possible)