If an element is reported as "Unmeasurable" in Table 7 or Status "-4" in a Score file, then ...

 

Observations for the element are part of an extreme score. One way is to impute non-extreme elements.

 

For instance:

 

Original specifications and data:

Models = ?, ?, R5 ; persons, items, rating-scale is 1-5

 

Data =

1, 1-2,  3, 4

2, 1-2,  3, 2

3, 1-2,  5, 5

4, 1-2.  1, 2

 

Person 3 is reported as "Unmeasurable". This is because it is the only person element rated 5, and all its ratings are 5. We know that 5 is better than 4, but the data do not tell us how much.

 

Let's say that 5 is only a little better than 4. Then we impute observations of 4 to Person 3:

 

Models =

3, ?, RScale, 0.5  ; weight observations of 3 at o.5

?, ?, RScale ; everyone else remains as before

*

Rating-scale = RScale, R5, General  ; Person 3 and everyone else share the same rating scale

 

Data =

1, 1-2,  3, 4

2, 1-2,  3, 2

3, 1-2,  5, 5  ; original data, now weighted 0.5

3, 1-2,  4, 4  ; imputed data, now weighted 0.5

4, 1-2.  1, 2

 

Person 3 will now be estimated with a measure corresponding to a rating of 4.5