Bias direction=, Write= and "B" in Models= statements control these Tables of bias interaction designators, "B", have been included in the measurement models, then bias estimates will be made. All estimates reported in the main analysis are fixed (anchored). Then Facets estimates the size of bias measures from the residuals of the main analysis. This process is repeated as many times as there are models with bias designators.
For each model specifying bias designators, a set of tables is produced, which estimate the designated bias as it is represented in all the modeled data (not just for the model containing the bias designators).
Table 9.1 Bias Iteration Report.
Bias/Interaction: 1. Senior scientists, 2. Junior Scientists
There are empirically 21 Bias terms
+-----------------------------------------------------------+
| Iteration Max. Score Residual Max. Logit Change |
| Elements % Categories Elements Cells |
|-----------------------------------------------------------|
| BIAS 1 -12.9867 -32.5 -1.1329 21 |
| BIAS 2 -1.1019 -2.8 -.1322 18 |
| BIAS 3 .0164 .0 -.0021 4 |
| BIAS 4 .0001 .0 .0000 0 |
+-----------------------------------------------------------+
Table 9
|
Report on the estimation of the size of the bias/interaction terms
|
9.1
|
the sequential number of the model specification that produced this set of Bias/Interaction Tables.
|
Bias/Interaction analysis specified by Model:
|
This shows the model statement.
|
There are empirically 21 Bias terms
|
the number of bias interaction terms for this model in the data. Suppose we have two facets selected for bias analysis, "B": items and ethnicities. Then Facets computes a bias size for every observed different combination of an item and an ethnicity. The maximum number of bias terms is (number of item elements) * (number of ethnicity elements). Empirically, not all these combinations may be observed, so the empirical number can be less.
|
Iteration =
|
how many times the data has been read to estimate bias terms.
|
Max. Score Residual:
|
the maximum (biggest) difference between an observed raw score and an expected score. 1.0 is the smallest observable difference. The convergence criterion is the minimum of Converge= and .01 score-points.
|
Elements
|
is the largest difference for any bias term.
|
Categories
|
(not relevant for bias terms)
|
Max. Logit Change:
|
is the largest change, in logits, between any bias estimate in this iteration and its estimate in the previous iteration. Starting estimates are 0.0 logits. The convergence criterion is the minimum of Converge= and .001 logits.
|
Elements
|
is the largest change for any bias term.
|
Cells
|
The number of Bias terms yet to be calculated to the specified convergence criteria
|
Both Max. Score Residual and Max. Logit Change should steadily reduce in absolute size, i.e., draw closer to zero. You may force termination by pressing the Ctrl+"F" keys simultaneously.
The detailed iteration report on your screen can be recorded in your output file with the "Write=Yes" specification.
