A two-step selective editing procedure based on contamination models

Ilves and Laitila (2009) propose a two-step procedure for selective editing. According to their approach, in addition to the units selected as affected by influential errors, a sample from the remaining observations is drawn in order to remove the possible residual bias. In this paper, the use of a contamination model as implemented in SeleMix (Buglielli and Guarnera, 2011) in the two-step procedure is studied. An evaluation study is performed by using data from 2008 Istat survey on small and medium enterprises and by simulating errors based on some mechanisms frequently met in NSI surveys.