SE (standard error is Model/Real) = Model

The "Model" standard error is the precision when of the Rasch measures when the data fit the model. The "Real" (enlarged) standard error increases the Model standard error to accord with the misfit in the data.

 

The Model SE is the smallest possible standard error (highest possible precision), and implies that the data accord with the measurement model, i.e, that misfit is random. If you hypothesize that noise in the data is systematic and not random (the worst case), and so wish to see standard errors (precisions) enlarged to encompass this systematic lack of fit discovered in the current analysis, then report the "Real" SE:

SE = Real ; "Real" can be abbreviated to "R", or specify SE=Enlarged

 

SE = Real is the more conservative approach. It is recommended for exploratory analysis.

Real enlarged S.E. = Model S.E. *  sqrt(Max(INFIT MnSq, 1))

 

When SE = Real, the reported Reliabilities in Table 7 are calculated using the Real standard errors. The "Real" reliabilities are lower than the default "Model" reliabilities,

 

The setting of SE= can be changed for post-hoc reporting in Modify Specifications.