One facet must be non-centered in most analyses, otherwise the estimates are over-constrained.
This specifies which facets are not to be centered. Other facets are centered, unless they are anchored. Using Noncenter=, centered facets are constrained to have a mean element measure of zero for non-extreme element measures. Center= can constrain facets to have a mean element measure of zero for all element measures.
In most analyses, if no facet is non-centered then the estimation is over-constrained and displacements are reported. If more than one facet is non-centered, the estimation is not sufficiently constrained and the estimates are ambiguous (non-identifiable). The usual convention is to center the frame of reference on the agents of measurement: items, tasks, judges. This establishes the origin. Then the objects of measurement, persons, are positioned with respect to this origin.
Extreme measures (corresponding to minimum and maximum possible scores) are included in the Means in Table 7, so that these may not be 0 for centered facets unless Center=(facet number)Z
If both Noncenter= and Center= are specified, only the first encountered is actioned.
When any element within a facet is anchored, or group-anchored, then the origin of the scale for that facet is set by the anchoring. Both Center= and Noncenter= are ignored.
Example 1: Persons are facet 2 and are not to be centered, then
Noncenter=2
Example 2: facet 1 contains the boys, and facet 3 the girls:
Noncenter=1,3
Example 3: All facets are anchored or centered or self-defined. Self-definition occurs when there is only one active facet so that the item difficulties are defined relative to an implicit origin, as in the "Three Mile" example.
Noncenter= or Noncenter=0
