The observations are understood to be the result of stochastic interactions between elements of facets. Each element is to be measured or calibrated within a common frame of reference. This frame of reference must be defined. There are three decisions:

 

a) Where is the scale origin is to be established for each facet? Center= or Noncenter=

Usually the origin of the scale, the "zero" point for each facet, is set at the mean value of the measures for that facet, i.e., centered. When this has been done for all facets except one, the placement of the origin for that last facet is then determined by the data. That facet will be non-centered=.

Usually the noncenter= facet is the examinees.

 

b) In which direction is each facet to be oriented? Positive= or Negative=

Does a greater raw score correspond to a greater measure? This positive orientation is usually how person abilities and judge leniencies are conceptualized.

Does a greater raw score correspond to a lesser measure? This negative orientation is usually how item and task difficulties and judge severities are conceptualized.

In some analyses, there is only one positively oriented facet, the measured people. All other facets are negatively oriented: item difficulty, judge severity, task challenge etc.

In other analysis, all facets are positively oriented. person ability, item easiness, judge leniency, task facility.

Careful choice of facet orientation avoids confusion in interpreting Facets output.

 

c) What is the unit of measurement? Umean=

The conventional Rasch unit of measurement is the "Logit", and this the standard unit. It is often convenient, however, to linearly rescale the logit into RITS, WITS, or to give the measures a useful range, e.g., 0 - 100 (see BTD chapter 8). Umean= accomplishes this.