Subset detection = Yes

Subsets in the data indicate a lack of identifiablity of the estimates. An infinity of different sets of estimates would produce the same fit to the Rasch model. Subsets = Yes requests Facets to report this problem of "connectedness".

Raw scores provide a Procrustean solution to the problem of connectedness: a rating of "1" implies the same level of performance everywhere, i.e, all judges are equally severe. Rasch says that the meaning of a "1" depends on its context. This enables more meaning to be extracted from the data, but also requires more care of the analyst and test designer. In Facets, Procrustean solutions are still available through the use of anchoring.

Subset detection = No or Bypass	do not detect subsets
Subset detection = Report	detect subsets and always show details in Table 6 and Table 7.
Subset detection = Yes (the default)	detect subsets, show details in Table 7, but do not show details in Table 6 unless checked in the Table 6 dialog box

Facets attempts to discover if the data permit the construction of one unambiguous measurement system. Specify Subset detect=No to bypass detection. Use this to speed up later runs, once data connectivity has been verified.

Subsets are listed in Table 6.0.0 and identified with the elements in Table 7. If subsets exist in the data, use the Output Files menu, Subset group-anchor file option.