Measures and calibrations

Various terms are used for the Rasch-model parameter estimates associated with each element.

 

Measures: this is the general term used to indicate the values of the estimates in a linear, interval, additive, frame of reference.

 

Logit values: the natural values for Rasch estimates are in "log-odds units" or logits. Logit differences have a direct relationship to probabilities for dichotomous items. For polytomous items, the probabilistic relationship is conditional.

 

Rescaled or user-scaled values: logit values can be rescaled linearly, i.e., by adding a constant amount and multiplying by a constant amount. The rescaled values maintain their linear quality. These are set with Umean=

 

Calibrations or difficulties: these are measures for items.

 

Abilities: these are measures for persons.

 

In dichotomous situations, the measure is the point on the latent variable at which there is a 50% probability of success or failure when encountering a person, item or other element known to be at that point on the variable.

 

In polytomous situations, the measure is the point on the latent variable at which there is an equal probability (however small) of being observed in the top or bottom polytomous category when encountering a person, item or other element known to be at that point on the variable.

 

The standard reference point for a dichotomy or polytomy can be altered by using structure anchoring, Rating scale=.