Anchor output file = " "

When Anchorfile= is specified, Facets writes out a new specification file containing the input specifications for this analysis. This file omits any specifications entered as extra specifications, but includes the final estimated measures for all elements and scales. This file also has all estimated measures marked as ",A" for "anchored".

 

These anchor values can be used as starting values for slightly different later runs. Edit the ",A"'s out of the first line of each facet in the Labels= specifications and out of the category lines of Rating (or partial credit) scale=.

 

This can also be specified from the Output Files menu, Anchor Output file.

 

Example: Using an edited version of Guilford.txt.

Anchorfile=Guilford.anc

 

Title = Ratings of Scientists (Psychometric Methods p.282 Guilford 1954)

Score file = GUILFSC        ; score files GUILFSC.1.txt, SC.2.txt and SC.3.txt produced

Facets = 3        ; three facets: judges, examinees, items

Inter-rater = 1        ; facet 1 is the rater facet

Arrange = m,2N,0f        ; arrange tables by measure-descending for all facets,

Positive = 2        ; the examinees have greater creativity with greater score

Non-centered = 1        ; examinees and items are centered on 0, judges are allowed to float

Unexpected = 2        ; report ratings if standardized residual >=|2|

Usort = (1,2,3),(3,1,2),(Z,3)        ; sort and report unexpected ratings several ways (1,2,3) is Senior, Junior, Trait

Vertical = 2N,3A,2*,1L,1A        ;define rulers to display and position facet elements

Zscore = 1,2        ;report biases greater in size than 1 logit or with z>2

Pt-biserial = measure ; point-measure correlation

Models =

?B,?B,?,RS1,1 ; CREATIVITY

*

Rating (or partial credit) scale = RS1,R9,G,O ; CREATIVITY

; Facets has renamed the rating scale in order to avoid ambiguity. Please edit the rating-scale name in the Anchorfile if you wish.

1=lowest,0,A

2=,-.6441868,A

3=,-2.317694,A

4=,.8300989,A

5=middle,-1.477257,A

6=,1.710222,A

7=,-1.001601,A

8=,2.358206,A

9=highest,.5422113,A

*

Labels =

1,Senior scientists,A

1=Avogadro,0.038439

2=Brahe,.2356183

3=Cavendish,-0.091576

*

2,Junior Scientists,A

1=Anne,-0.068648

2=Betty,.637455

3=Chris,-.2451435

4=David,-.4621402

5=Edward,.4228902

6=Fred,-.5607015

7=George,.2762882

*

3,Traits,A

1=Attack,-.2684814

2=Basis,-.1409857

3=Clarity,.2020537

4=Daring,-.2898233

5=Enthusiasm,.4972367

*

Data=

1,1,1-5,5,5,3,5,3

....

 

Example 1: Two time points and a rating scale:

 

(1) Obtain the common structure for the rating scale:

Construct the entire data set. Put in a dummy "time-point" facet with the two elements anchored at 0.

If the same person appears twice, then give them two related id-numbers, so that they are easy to pair up later. Analyze the dataset. Write out an Anchorfile=

 

(2) Time-point 1 is the "gold standard":

For the time-1 data, edit the Model= statements for element 1 of the time-point facet.

Copy-and-paste into the specification file the parts you want to anchor from the (1) anchor file.

Analyze the time-1 data and output another anchorfile.

 

(3)  Time-point 2 is measured in the Time-point 1 "frame of reference":

For the time-2 data, edit the Model= statements for element 2 of the time-point facet.

Copy-and-paste into the specification file the parts you want to anchor from the (2) anchor file.

Analyze the time-2 data.

 

Example 2: I want generate a report for each criterion entered in the analysis. I am interested in getting fair averages for the different criteria used by the raters.

1. Perform the complete analysis.

2. Output an Anchorfile=

3. In the Anchorfile=, keep everything anchored but comment out all except one criterion.

4. Analyze the Anchorfile=.

Now all the reporting will be for only that one criterion.

5. Return to 3. for the next criterion.

 

Example 3: Fair Average score for Item-by-Candidate combinations.

i.Do a standard analysis:
Facets = 3 ; Candidates, Raters, Items
Non-centered = 1 ; candidates - we need this
Model = ?, ?, ?, R
Anchorfile = anc.txt ; we need this

ii.In anc.txt,
Edit the model specification:
Model = ?, X, ?,  ....   ; inactivates the raters
Remove "Anchorfile="
Residual file = residual.xls ; Excel file - we need this

iii.Analyze anc.txt

iv.Open Residual.xls ; in Excel
Delete all columns except: Candidate, Item, Expected score
Sort the file: Candidate major, Item minor.
All the expected scores for each candidate-item combination should be the same
Used the Excel "Advanced filter" to remove duplicate lines.

v.The "expected score" is the  "Fair Average score for Item-by-Candidate"