Statistical biases in measurements with multiple candidates

Many measurements at collider experiments study physics candidates that are a subset of a collision event. The presence of multiple such candidates in a given event can cause raw biases which are large compared to typical statistical uncertainties. Selecting a single candidate is common practice but only helps if the likelihood of selecting the true candidate is very high. Otherwise the precision of the measurement can be affected, and biases can be generated, even if none are present in the data sample prior to this operation. This paper aims at describing the problem in a systematic way. It sets definitions, provides examples of potential biases using pseudoexperiments and gives recommendations.
Comment: This is the 2019 update of the original 2017 preprint