Bias
A survey’s answer can be wrong in the same direction every time it is run. Where does that lean come from?
Where you are 3 inside
What this area is about
A sample never matches its population exactly, and chance error goes up one time and down the next. Bias is different: a way of collecting data that keeps favoring some answers, so its results lean the same way every time it is used. A bigger sample does not cure it. This area names where the lean comes from, so that from a study’s description you can say who was missed or how answers were bent, and which way that pushes the result.
How these fit together
The three children are three places a lean can enter.
Who is chosen. Convenience and voluntary response samples let the collector, or the respondents themselves, decide who is in the sample.
Who could be chosen, and who replies. Undercoverage and nonresponse are the gaps a chance draw cannot close: people missing from the list it draws from, and people drawn who never answer, with the response rate \displaystyle \frac{r}{n} to measure the second.
How they answer. Response bias comes from the answers themselves: sensitive questions, wanting to please, poor memory, and leading or loaded wording.
A path through it
The three can be taken in any order; undercoverage reads best after random sampling, since it is about the gaps a random draw leaves.
Finish here and you can judge how data were gathered before trusting them, which observational studies and experiments builds on.