Selection Bias vs Nonresponse Bias
Both terms below come up in the same part of the course, and students mix them up. Here is each one defined on its own, side by side, so you can see where they part company.
Selection bias
Collecting data and study design
Selection bias is bias created by how the sample was chosen, leaving the individuals studied systematically different from the population.
Any method that gives some groups a better chance of being included than others can push a statistic consistently above or below the parameter. Convenience samples, voluntary response samples, and a sampling frame with undercoverage are the usual ways it gets in. For example, surveying shoppers at a store on a Tuesday morning overrepresents people who are not at work on weekday mornings, so an estimate of average income can come out too low. Because the tilt is systematic rather than random, a bigger sample makes the wrong answer more precise instead of more correct.
Nonresponse bias
Collecting data and study design
Nonresponse bias occurs when people chosen for the sample do not respond, and those who do respond differ systematically from those who do not.
Nonresponse bias arises after the sample is chosen, when some selected people do not answer. For example, if a survey reaches a random sample but only the most satisfied customers reply, the results skew positive. The problem is not the missing answers themselves but that non-responders tend to differ from responders. Unlike undercoverage, which happens before contact, nonresponse occurs after people have been selected.