Sample Selection Bias

When we select a sample from a larger population, we can screw up taking the sample in any number of ways. (No xx/ relation to candy, urine, stool or blood sampling here.) The ways that we take bad samples all typically introduce sample selection bias which happens when we systematically don’t select a sample that is representative of how our population itself is composed.

Let’s say we worked for a high school newspaper and decided to do some hard-hitting journalism on if there’s an issue finding parking places in the school lot in the morning. School starts at 8:00 am, so we head out at 6:50 am to grab our sample of the first 40 people arriving in the lot. Strangely, all 40 of those people say, “No, there’s plenty of spaces. Are you mental?”

Our sample is loaded with sample selection bias because we ignored the people who get there closer to when school starts when there might be fewer free spaces. Ideally our sample should always reflect a similar breakdown of individuals that the population does in terms of issues that might affect our results. In the case of the article, we should have a sample that has some early risers, some on-timers, and some of the kids who are running in as the late bell is ringing.