PAFILE= person anchor file

The PFILE= from one run can be used unedited as the person anchor file, PAFILE=, of another.

 

 

The person parameter values (thetas) can be anchored (fixed) using PAFILE=. Person anchoring can also facilitate test form equating. The persons common to two test forms can be anchored at the values for one form. Then the measures constructed from the second form will be equated to the measures of the first form. Other measures are estimated in the frame of reference defined by the anchor 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=.

 

PAFILE=? opens a Browse window

 

In order to anchor persons, a data file must be created of the following form:

1. Use one line per person (or person range) to be anchored.

2. Type the sequence number of the person in the current analysis, a blank, and the measure-value at which to anchor the person (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 person appears more than once, the first anchor value is used.

 

UIMEAN= and UPMEAN= are ignored when there are anchor values, IAFILE= or PAFILE=

 

Examples:

PAFILE=*

2 3.47 ; anchors person 2 at 3.47 logits (or USCALE= values)

10-13 1.3 ; persons 10, 11, 12, 13 are each anchored at 1.3 logits

2 5.2 ; person 2 is already anchored. This person anchoring is ignored

1-50 0 ; all the unanchored persons in the range 1-50 are anchored at 0.

*

Anything after ";" is treated as a comment.

 

PAFILE = filename

Person anchor information is in a file containing lines of format

 person entry number       anchor value

 person entry number       anchor value

 

PAFILE=*

Person anchor information is in the control file in the format

 PAFILE=*

 person entry number     anchor value

 person entry number     anchor value

 *

 

PAFILE=$SnnEnn or PAFILE=$SnnWnn

Person anchor information is in the person data records using the column selection rules, e.g., starting in column Snn and ending in column Enn or of width Wnn. Blanks of non-numeric values indicate no anchor value. PAFILE=$S10E12 or PAFILE=$S10W2 means anchor information starts in column 10 and ends in column 12 of the person's data record (not person label). This can be expanded, e.g, PAFILE = $S23W1+"."+$S25W2 places the columns next to each other (not added to each other)

 

Example 1: The first 3 persons are to be anchored at pre-set values

TITLE = "PAFILE TEST"

ITEM1 = 1

NI = 4

NAME1 = 6

CODES = 01

PAFILE=*

1 0.6  ; PERSON 1 ANCHORED AT 0.6 LOGITS

2 2.4 ; PERSON 2 ANCHORED AT 2.4 LOGITS

3 -1.8 ; PERSON 3 ANCHORED AT -1.8 LOGITS

*

&END

Item 1

Item 2

Item 3

Item 4

END LABELS

0101 Person 1

1110 Person 2

1000 Person 3

0110 Person 4

 

Example 2: The third person in the test is to be anchored at 1.5 logits, and the eighth at -2.7.

1. Create a file named, say, "PERSON.ANC"

2. Enter the line "3 1.5" into this file, meaning "person 3 is fixed at 1.5 logits".

3. Enter the line "8 -2.7", meaning "person 8 is fixed at -2.7 logits".

4. Specify, in the control file,

PAFILE=PERSON.ANC

CONVERGE=L  ; only logit change is used for convergence

LCONV=0.005  ; logit change too small to appear on any report.

 

or, enter directly into the control file

 PAFILE=*

 3 1.5

 8 -2.7

 *

 CONVERGE=L  ; only logit change is used for convergence

 LCONV=0.005  ; logit change too small to appear on any report.

 

or, include the anchor information in the data record

 PAFILE=$S1E4

 NAME1=5

 ITEM1=11

 NI=12

 CONVERGE=L  ; only logit change is used for convergence

 LCONV=0.005  ; logit change too small to appear on any report.

 &END

 .....

 END LABELS

     Fred  111010111001 ; this is the first data line

     Mary  011010111100

 1.5 Jose  101001111010 ; this data line has an PAFILE= anchor value

     Jo    111110011101

etc.

 

To check: "A" after the measure means "anchored"

 

+----------------------------------------------------------------------------------------+

|ENTRY    RAW                        |   INFIT  |  OUTFIT  |PTMEA|        |              |

|NUMBER  SCORE  COUNT  MEASURE  ERROR|MNSQ  ZSTD|MNSQ  ZSTD|CORR.|DISPLACE| PERSONS      |

|------------------------------------+----------+----------+-----+--------+--------------|

|     3     32     35     1.5A    .05| .80   -.3| .32    .6|  .53|     .40| Jose         |

 

Example 2: Matching person labels.

Winsteps expects the person entry numbers for the common persons to be the same in both tests. If they are not, then here is a procedure using Excel.

 

Run the two analyses independently.

Output the two PFILE= to Excel files.

Sort the two Excel Pfiles by person label.

Copy and paste the two files into one excel worksheet:

 

The columns are:

Columns A,B,C: Test 1: person id, entry number, measure

Columns D,E :Test 2:  person id, entry number

 

The person labels will not align, because some people are in Test 1 and not Test 2, and some people are in Test 2 but not Test 1.

 

We need to apply anchor values from Test 1 in the Test 2 analysis.

 

Now we want Excel to do a look-up for us:

Paste into cell F1:

=VLOOKUP(D1,A$1:C$999,3,FALSE)

This looks up the person label in cell D1 in column A, and puts the measure in column C into column F1.

Change 999 to match the number of rows in column A

 

Copy and paste cell F1 into F2:F999 (to the bottom row of Test 2)

The anchor measures are now in column F, with #N/A where there is no measure.

 

Copy and paste columns E and F into a new text file: test2pafile.txt.

Add to your Test 2 control file:

PAFILE=test2pafile.txt

The lines containing #N/A will be ignored by Winsteps.

Perform the Test 2 analysis.

Check that all is correct by looking at the entry-order measure Table for Test 2.