Tracing selection effects in three non-probability samples.

Snowball sampling and targeted sampling are widely applied techniques to recruit samples from hidden populations, such as problematic drug users. The disadvantage is that they yield non-probability samples which cannot be generalised to the population. Despite thorough preparatory mapping procedures, selection effects continue to occur. This paper proposes an interpretation frame that allows estimating the direction of selection bias after data collection. Critical examination of the recruitment procedure and comparison with statistical and non-statistical external data sources are the core features of the interpretation frame. Applying the interpretation frame increases insight into the reliability of the results and allows to estimate where selection bias may have occurred.