The size of the estimated bias in Table 12 and Table 13 is reported in logits(or Umean= units), and its significance is reported as a t-statistic. The direction, i.e., sign, of the bias depends on how it is to be interpreted. In Table 14 and on the Excel plots, the bias direction accords with the reported facet direction, Positive=.

 

In the specific observations being analyzed for local bias:

Bias Direction = Plus or + or Positive or Ability or Leniency or Easiness (higher score = higher measure)

Bias Direction = Minus or - or Negative or Difficulty or Severity (higher score = lower measure)

 

Observed score > Expected score

Person is more able here than overall. Element is less difficult here than overall

 

Observed score < Expected score

Person is less able here than overall. Element is more difficult here than overall

 

Mathematically:

local measure = global measure + bias measure

local measure = global measure - bias measure

 

To suppress bias analysis: Bias direction = Omit

 

Example: Table 13:

Bias =  Plus

-----------------------------------

|Obsvd Exp.  Obsvd Obs-Exp| Bias

|Score Score Count Average| Measure

----------------------------------

| 9791 10506.0 6771 -.11 | -.10 <= observed-expected is negative (lower person performance than expected), so person ability bias is negative.

 

Bias = Minus

-----------------------------------

|Obsvd Exp.  Obsvd Obs-Exp| Bias

|Score Score Count Average| Measure

----------------------------------

| 9791 10506.0 6771  -.11 |  .10 <= observed-expected is negative (higher difficulty than expected), so item difficulty bias is positive.