Solange Koch (National Educational Panel Study )
In general sampling is based on a complete list. Complete lists of information concerning all units of the target population enable using sampling strategies and getting – depending on the quality of list and strategy – unbiased estimators for parameters of the target population. Without a sampling frame direct sampling is not possible. In this situation Lavallée (2007) suggests to use an indirect sampling approach. If there exists a population with an available sampling frame and this population is connected to the target population in a uniquely defined manner, indirect sampling allows provision of a random sample of the target population and the provision of weights permitting an unbiased estimation of population parameters. There are different kinds of indirect sampling approaches, depending on the kind of link between two populations UA (the population with an available sampling frame) and UB (the target population) and whether a multistage sampling strategy is incorporated. The connection between both populations is called link and can be described by an indicator variable: If there is a connection between a unit j of UA and a unit k of UB , then let lj,ik =1 and lj,ik =0 otherwise, with i describing the structure of clusters in population UB , if there exists one .With this indicator variable and the knowledge of the selection probabilities of all units in UA weights can be calculated.
In the National Educational Panel Study (NEPS) six main cohorts are sampled by a multicohort sequence design to get a longitudinal analysis of education in Germany. In contrast to institutions of other cohorts there is no sampling frame available for Kindergartens. However, in Germany all children have to attend at elementary school. Therefore every child that had been in Kindergarten can be found in an elementary school and for elementary schools a sampling frame exists. Hence, in the NEPS we sample elementary schools sA from the population UA and ask them to list all Kindergartens k from which children from their first grade came from. Proportional to the number of trespassing children Kindergartens are sampled and Kindergarten children tested. With all children tested in Kindergarten the sampling procedure is a two-stage indirect sampling.