Unexpected observations (with standardized residuals greater than Unexpected=) are listed in Table 4. Usort= controls the order in which the unexpected observations are listed.
Usort=
|
facet number: 1,2,3,..
|
Sort residuals by this facet location, ascending
|
,
|
separator between facet numbers and lists
|
( )
|
sort and list delimiters
|
|
Sort order:
|
|
Ascending
|
Descending
|
Sorted by:
|
C
|
c
|
Category or Data or Observation, i.e., the original scored response (also D, O)
|
X
|
x
|
Step Value or Xni, i.e., the recounted value used for estimation (also V)
|
E
|
e
|
Expected value or Pni, i.e., the modeled value computed from the measures (also P)
|
R
|
r
|
Residual or Yni, i.e., the difference between the step value and the modeled value, Yni = Xni - Pni (also Y)
|
Z
|
z
|
Standardized residual or Zni, i.e, the statistical unexpectedness of the observation (also S, F, T)
|
U
|
u
|
Unexpectedness: Absolute value of standardized residual
|
Example 1: In Table 4, it is often useful to sort the unexpected residuals by their facet location and size. Sort the residuals by Facet 1 major, Facet 2 minor and then Facet 3.
Usort=1,2,3
Multiple sorts and multiple listings of the residuals are obtained by placing each sort order in (). Produces two listings, one sorted by Facet 2 major, the other sorted by Facet 3.
Usort=(2,1,3),(3,2,1)
Example 2: Persons are facet 1, items are facet 2 and judges are facet 3. We want to investigate the judges one at a time, paying attention to the most unexpected residuals for each item in Table 4, then identify grossly unexpected person ratings:
Usort = (3,2,u),(u,1)
Unexpected= 3 ; only for very unexpected ratings
