STAT 778 PARSCALE TEMPLATES

Stat 778 - Spring 2006 - PARSCALE Templates

The Basics of PARSCALE:

The basic example for PARSCALE uses the 8 item polytomous data set referred to in homework 3 question 7 and created using the code that can be found in the R templates. Each of the items is scored from 0 to 4 in this example and the data set is assumed to be saved as polymath.dat on the Z drive.

Before using PARSCALE it is very helpful to have a sample .PSL command file, as PARSCALE does not have menu options for constructing one from scratch. As in BILOG-MG, when starting PARSCALE for the first time you will be confronted with a gray screen. If you already have a saved .PSL file you can use File > Open to find it. If not, you can choose FILE > NEW and choose the directory you want to save the .PSL file in.

The lines below are a sample .PSL file for anlyzing polymath.dat.

Sample Parscale Analysis 
of polymath.dat
>COMMENT
This data set was first mentioned on homework 3
>FILE    DFNAME='Z:/POLYMATH.DAT',SAVE;
>SAVE    PARM='Z:/POLYMATH.PAR',SCORE='Z:/POLYMATH.SCO';
>INPUT   NIDCHAR=5,NTOTAL=8,NTEST=1,LENGTH=8;
(5A1,1X,8A1)
>TEST    ITEM=(1(1)8),NBLOCK=8;
>BLOCK1  NITEMS=1,NCAT=5,ORIGINAL=(0,1,2,3,4);
>BLOCK2  NITEMS=1,NCAT=5,ORIGINAL=(0,1,2,3,4);
>BLOCK3  NITEMS=1,NCAT=5,ORIGINAL=(0,1,2,3,4);
>BLOCK4  NITEMS=1,NCAT=5,ORIGINAL=(0,1,2,3,4);
>BLOCK5  NITEMS=1,NCAT=5,ORIGINAL=(0,1,2,3,4);
>BLOCK6  NITEMS=1,NCAT=5,ORIGINAL=(0,1,2,3,4);
>BLOCK7  NITEMS=1,NCAT=5,ORIGINAL=(0,1,2,3,4);
>BLOCK8  NITEMS=1,NCAT=5,ORIGINAL=(0,1,2,3,4);
>CAL     PARTIAL,LOGISTIC,NQPTS=40,CYCLES=(100,2,2,2,2,1),
         CRIT=0.001,NEWTON=5,SPRIOR,TPRIOR;
>SCORE   EAP,DIST=2,ITERATION=(0.001,40),NQPT=40;

Simply choose Run > Run All to produce the output.

To modify this to analyze another data set, begin by choosing File > Save As and choosing a name for the new PSL file. The key things to change are:

The title (the first two lines)
The comment given below COMMENT
The data file name DFNAME and location
The file names and locations for the parameter estimates (PARM) and the ability estimates (SCORE)
The number of ID characters (the NIDCHAR=5 and 5A1 on the next line)
The number of items (the NTOTAL=8, LENGTH=8, 8A1, ITEM=(1(1)8), and NBLOCK=8
Each item has its own block where NCAT is the number of possible responses, and those responses are given in the original vectors.

Once the changes are made you can File > Save and Run > Run All.

The item parameters are stored in the .PH2 file in two parts, the first looks something like:

   ITEM BLOCK   1  BLOCK   

   SCORING FUNCTION    :     1.000     2.000     3.000     4.000     5.000

   STEP PARAMTER       :     0.000     1.274     0.213    -0.204    -1.282

   S.E.                :     0.000     0.119     0.081     0.066     0.060

and are the d_h estimates using Muraki's notation. The second part of the item parameter estimates looks something like:

+------+-----+---------+---------+---------+---------+---------+---------+
| ITEM |BLOCK|  SLOPE  |   S.E.  |LOCATION |   S.E.  |GUESSING |   S.E.  |
+======+=====+=========+=========+=========+=========+=========+=========+
| 0001 |   1 |   0.515 |   0.023 |  -0.873 |   0.035 |   0.000 |   0.000 |
+------+-----+---------+---------+---------+---------+---------+---------+

and gives the estimates of the a_i and b_i parameters. As parscale can also estimate the 3PL model for dichotomous items space for the guessing parameters is also given.

Notice that these same values can be found without column labels in the .PAR file generated by the program.

The examinee ability estimates can be found in the .PH3 file, or in the .SCO file without the additional information.

It should be noted that PARSCALE can also estimate the 3PL model or a combination of Partial Credit and 3PL items. The following code will analyze the mdatabid.dat data used in the BILOG example. Notice that the two programs apparently estimate the parameters somewhat differently and the examinee ability estimates need to be put on a common scale before comparing them.

Sample Parscale Analysis 
of mdatabid.dat
>COMMENT
This is the data set analyzed in R and with BILOG
>FILE    DFNAME='Z:/MDATABID.DAT',SAVE;
>SAVE    PARM='Z:/MDATABIDPARS.PAR',SCORE='Z:/MDATABIDPARS.SCO';
>INPUT   NIDCHAR=5,NTOTAL=32,NTEST=1,LENGTH=32;
(5A1,1X,32A1)
>TEST    ITEM=(1(1)32),NBLOCK=1;
>BLOCK  NITEMS=32,NCAT=2,ORIGINAL=(0,1),GUESSING=(2,ESTIMATE);
>CAL     PARTIAL,LOGISTIC,NQPTS=40,CYCLES=(40,40,40,40,40,1),
         CRIT=0.001,NEWTON=5,SPRIOR,TPRIOR,GPRIOR;
>SCORE   EAP,DIST=2,ITERATION=(0.001,40),NQPT=40;