IAFILE= item anchor file |
The IFILE= from one analysis can be used unedited as the item anchor file, IAFILE=, of another.
IAFILE= *file name |
file containing details |
IAFILE = * |
in-line list |
IAFILE = $S1W1 |
field in item label |
IAFILE=? |
opens a Browser window to find a file containing the details |
The item parameter values (deltas) can be anchored (fixed) using IAFILE=. Anchoring facilitates equating test forms and building item banks. The items common to two test forms, or in the item bank and also in the current form, can be anchored at their other form or bank calibrations. Then the measures constructed from the current data will be equated to the measures of the other form or bank. Other measures are estimated in the frame of reference defined by the anchor values. The anchored values are imputed (inserted) in place of the estimated-from-the-data values. Mathematically, the anchor values are treated as though they, like the estimated-from-the-data values, are the best available estimates of the true values.
Displacements are reported, indicating the differences between the anchored values and the freely estimated values. If these are large, please try changing the setting of ANCESTIM=.
For polytomies (rating scales, partial credit), IAFILE= must have SAFILE=. The IFILE= and the SFILE= are really one file. For dichotomies, the SFILE= is uninformative, so it can be ignored. For polytomies, the IFILE= and the SFILE= form a pair, and so do the IAFILE= and the SAFILE=. For polytomies, anchoring with the IAFILE= without the SAFILE= is usually meaningless. The items are not completely anchored. Use IAFILE= and SAFILE= if you need the polytomous item in one analysis to be identical in thresholds and overall difficulty to the same item in another analysis. Use only SAFILE= if you need the polytomous item in one analysis to be identical in thresholds to the same item in another analysis, but the overall item difficulties can differ.
How anchoring works:
The anchored items together with the unanchored items determine the person measures based on the data. The person measures determine the calibrations of the unanchored items and the displacements of the anchored items. The person measures are adjusted so that the mean displacement of the anchored items is zero.
Let's imagine some situations with complete data:
1. The data fit the anchored anchored items exactly. There are no displacements and the unanchored items slot exactly into the hierarchy of the anchored items. Person measures are the same as in an unanchored analysis, but the mean ability measure is adjusted so that the anchored item displacements are all zero. Unanchored items with the same p-values as the anchored items have the same calibrations.
2. All the anchored items happen to have the same item calibration, but have different p-values in the data. The mean ability measure is adjusted so that the mean anchored item displacement is zero. The ability measures are more central than in an unanchored analysis. The calibrations of the unanchored items are more central than in an unanchored analysis, but not the same as anchored items with the same p-values.
3. The anchored items have calibrations that are random with respect to the current data. The mean ability measure is adjusted so that the mean anchored item displacement is zero. The ability measures are more central than in an unanchored analysis. The calibrations of the unanchored items are more central than in an unanchored analysis, but not the same as anchored items with the same p-values.
4. The anchored items have calibrations that are correlated with the current data, but more extreme than their values in an unanchored analysis. The mean ability measure is adjusted so that the mean anchored item displacement is zero. The ability measures are more diverse than in an unanchored analysis. The calibrations of the unanchored items are more diverse than in an unanchored analysis, but not the same as anchored items with the same p-values.
Anchor file format:
In order to anchor items, a data file must be created of the following form:
1. Use one line per item (or item range) to be anchored.
2. Type the sequence number of the item in the current analysis, a blank, and the measure-value at which to anchor the item (in logits if UASCALE=1, or in your user-rescaled USCALE= units otherwise). Arithmetical expressions are allowed.
Further values in each line are ignored. An IFILE= works well as an IAFILE=.
3. If the same item appears more than once, the first anchor value is used. When an IFILE= will be used as an IAFILE=, be sure to output the measures with many decimal places: UDECIMALS=4
UIMEAN= and UPMEAN= are ignored when there are anchor values, IAFILE= or PAFILE=
Stopping estimation: usually Winsteps estimation converges successfully by itself. If it does not, the ctrl+F stops estimation. If this happens repeatedly for an analysis, then you can explicitly tell Winsteps to do whatever you see when you decide to end estimation. For instance, if your decide to stop estimation when the biggest change to logit estimates is less than .01 logits,
LCONV=.01
With anchor values, which usually mean that sum score residuals will never be zero, this choice makes sense.
Examples:
2 3.47 ; anchors item 2 at 3.47 logits (or USCALE= values)
10-13 1.3 ; items 10, 11, 12, 13 are each anchored at 1.3 logits
2 5.2 ; item 2 is already anchored. This item anchoring is ignored
1-50 0 ; all the unanchored items in the range 1-50 are anchored at 0.
Anything after ";" is treated as a comment.
IAFILE = filename
Item anchor information is in a file containing lines of format
item entry number anchor value
item entry number anchor value
IAFILE=*
Item anchor information is in the control file in the format
IAFILE=*
item entry number anchor value
item entry number anchor value
*
IAFILE=$SnnEnn or IAFILE=$SnnWnn or @Field
Item anchor information is in the item labels using the column selection rules. Blanks or non-numeric values indicate no anchor value.
Anchoring and Extreme items
In the original calibration, extreme items are given an estimated finite difficulty. They are not used in fit reporting and person measurement.
In the anchored calibration, all items are anchored at their estimated difficulties, including the previously extreme items. All items including previously extreme items are used in fit reporting and person measurement.
If the anchor values for the extreme items are not to be used in the anchored analysis, we need to eliminate the extreme items from the anchor file:
1. In an interactive run of Winsteps,
1. analyze any dataset
2. Output Files menu
3. IFILE=
4. Select fields and other options
5. Check "flag extremes with ;"
6. "Make default"
7. Cancel your way out of Winsteps
The result is that all IFILE= output files will have ; as the first character of extreme score lines These will be treated as comments when processed as anchor files by IAFILE=.
Example 0: only one item is to be anchored:
Slow method - include in your control file:
USCALE=1 ; anchor value and analysis in logits
CONVERGE= L ; Convergence decided by logit change
LCONVERGE=.00001 ; Set logit convergence tight because of anchoring
IAFILE = * ; Item anchor file to preset the difficulty of an item
6 0.25 ; Item 6 exactly at 0.25 logit point.
*
Faster method:
1) do a standard unanchored analysis
2) output Table 14 items
3) see the measure for item 6 (for me it is1.30)
4) edit your control file so that UIMEAN = wanted value - current value = 1.30 - 0.25 = 1.05
5) do the standard unanchored analysis again: item 6 is now 0.25
Example 1: The third item is to be anchored at 1.5 logits, and the fourth at 2.3 logits.
1. Create a file named, say, "ANC.FIL"
2. Enter the line "3 1.5" into this file, which means "item 3 in this test is to be fixed at 1.5 logits".
3. Enter a second line "4 2.3" into this file, which means "item 4 in this test is to be fixed at 2.3 logits".
3. Specify, in the control file,
USCALE=1 ; anchor value and analysis in logits
IAFILE=ANC.FIL
CONVERGE=L ; only logit change is used for convergence
LCONV=0.005 ; logit change too small to appear on any report.
or place directly in the control file:
IAFILE=*
3 1.5
4 2.3
*
CONVERGE=L ; only logit change is used for convergence
LCONV=0.005 ; logit change too small to appear on any report.
or in with the item labels:
IAFILE=$S10W4 ; location of anchor value in item label
CONVERGE=L ; only logit change is used for convergence
LCONV=0.005 ; logit change too small to appear on any report.
&END
Zoo
House 1.5 ; item label and anchor value
Garden 2.3
Park
END LABELS
To check: "A" after the measure means "anchored"
+----------------------------------------------------------------------------------------+
|ENTRY RAW | INFIT | OUTFIT |PTMEA| | |
|NUMBER SCORE COUNT MEASURE ERROR|MNSQ ZSTD|MNSQ ZSTD|CORR.|DISPLACE| ITEMS |
|------------------------------------+----------+----------+-----+--------+--------------|
| 3 32 35 1.5A .05| .80 -.3| .32 .6| .53| .40| House |
Example 2: The calibrations from one run are to be used to anchor subsequent runs. The items have the same numbers in both runs. This is convenient for generating tables not previously requested.
1. Perform the calibration run, say,
C:> Winsteps SF.TXT SOMEO.TXT IFILE=ANCHORS.SF TABLES=111
2. Perform the anchored runs, say,
C:> Winsteps SF.TXT MOREO.TXT IAFILE=ANCHORS.SF TABLES=0001111
C:> Winsteps SF.TXT CURVESO.TXT IAFILE=ANCHORS.SF CURVES=111
Example 3: Score-to-measure Table 20 is to be produced from known item and rating scale structure difficulties.
Specify:
USCALE=values ; scaling of the anchor values
IAFILE=iafile.txt ; the item anchor file
SAFILE=safile.txt ; the structure/step anchor file (only for polytomies)
TFILE=*
20 ; the score table
*
CONVERGE=L ; only logit change is used for convergence
LCONV=0.005 ; logit change too small to appear on any report.
STBIAS=NO ; anchor values do not need estimation bias correction.
The data file comprises two dummy data records, so that every item has a non extreme score, e.g.,
For dichotomies:
CODES = 01
Record 1: 10101010101
Record 2: 01010101010
For a rating scale from 1 to 5:
CODES = 12345
Record 1: 15151515151
Record 2: 51515151515
Example 4. Anchoring polytomous items for the Rating Scale Model
CODES = 012 ; 3 category Rating Scale Model
IAFILE=*
1 2.37 ; anchor item 1 at 2.37 logits
2 -1.23
*
SAFILE=*
0 0 ; the bottom category is always anchored at 0
1 -2.34 ; Andrich threshold (step difficulty) from category 0 to 1
2 2.34 ; Andrich threshold (step difficulty) from category 2 to 3
*
Example 5. Anchoring polytomous items for the Partial Credit and Grouped-Items models
CODES = 012 ; 3 category Rating Scale Model
ISGROUPS=0
IAFILE=*
1 2.37 ; anchor item 1 at 2.37 logits
2 -1.23
*
SAFILE=*
; for item 1, relative to the difficulty of item 1
1 0 0 ; the bottom category is always anchored at 0
1 1 -2.34 ; Andrich threshold (step difficulty) from category 0 to 1
1 2 2.34 ; Andrich threshold (step difficulty) from category 2 to 3
; for item 2, relative to the difficulty of item 2
2 0 0 ; the bottom category is always anchored at 0
2 1 -1.54 ; Andrich threshold (step difficulty) from category 0 to 1
2 2 1.54 ; Andrich threshold (step difficulty) from category 2 to 3
*
Help for Winsteps Rasch Measurement and Rasch Analysis Software: www.winsteps.com. Author: John Michael Linacre
Facets Rasch measurement software.
Buy for $149. & site licenses.
Freeware student/evaluation Minifac download Winsteps Rasch measurement software. Buy for $149. & site licenses. Freeware student/evaluation Ministep download |
---|
Forum: | Rasch Measurement Forum to discuss any Rasch-related topic |
---|
Questions, Suggestions? Want to update Winsteps or Facets? Please email Mike Linacre, author of Winsteps mike@winsteps.com |
---|
State-of-the-art : single-user and site licenses : free student/evaluation versions : download immediately : instructional PDFs : user forum : assistance by email : bugs fixed fast : free update eligibility : backwards compatible : money back if not satisfied Rasch, Winsteps, Facets online Tutorials |
---|
Coming Rasch-related Events: Winsteps and Facets | |
---|---|
Oct 21 - 22 2024, Mon.-Tues. | In person workshop: Facets and Winsteps in expert judgement test validity - UNAM (México) y Universidad Católica de Colombia. capardo@ucatolica.edu.co, benildegar@gmail.com |
Oct. 4 - Nov. 8, 2024, Fri.-Fri. | On-line workshop: Rasch Measurement - Core Topics (E. Smith, Winsteps), www.statistics.com |
Jan. 17 - Feb. 21, 2025, Fri.-Fri. | On-line workshop: Rasch Measurement - Core Topics (E. Smith, Winsteps), www.statistics.com |
May 16 - June 20, 2025, Fri.-Fri. | On-line workshop: Rasch Measurement - Core Topics (E. Smith, Winsteps), www.statistics.com |
June 20 - July 18, 2025, Fri.-Fri. | On-line workshop: Rasch Measurement - Further Topics (E. Smith, Facets), www.statistics.com |
Oct. 3 - Nov. 7, 2025, Fri.-Fri. | On-line workshop: Rasch Measurement - Core Topics (E. Smith, Winsteps), www.statistics.com |
Our current URL is www.winsteps.com
Winsteps® is a registered trademark