Both BILOG-MG and PARSCALE are best run from a DOS prompt in the directory where you want to store all of your data This does not have to be the same directory where you have installed the programs.
If you do not know how to find the DOS prompt on your machine, try using the Help option under the Start menu. In Windows 2000, you can search for Command Prompt and under that you should see finding MS-DOS Prompt in Windows.
When you start the DOS prompt up you should get a black window with greyish writing. There are just a few commands you need to know in order to get to the correct directory. I am using the directory 778 on my d-drive. At the prompt (">") in the DOS window I would simply type:
d: cd 778Typing the first line and hitting return selects my d-drive, and typing the second chooses the directory (hence cd) 778. You, of course, might be saving it on your C drive in some other directory.
This is the directory where I am going to run all of my analyses from, and save all of the results to. In particular, you need to save the data set and the command files in this directory. The command files for BILOG end in .BLM and those for PARSCALE end in .PSL. Examples of such a set of files are: http://www.stat.sc.edu/~habing/courses/simul.dat, http://www.stat.sc.edu/~habing/courses/data/bilogex.BLM, and http://www.stat.sc.edu/~habing/courses/data/parsex.PSL.
If I wanted to look at any of these files I would not need to do so in DOS. Instead, I would probably look at them using the program NOTEPAD, that you can find under Accessories in the Programs sub-menu of the Start Menu. Notice that the two command files both have several places that begin with D:/778/. Make sure and change that to the drive and folder that you are using!
Once I am sure the command files are correct, and that I am in the correct directory in the DOS prompt, I would run BILOG simply by typing
bilogmg bilogexand would run PARSCALE by typing
parscale parsexIf they run correctly, various computations should flash across the screen, and a variety of output files should appear in your directory. Note that you do not include the extensions .BLM and .PSL when you are typing the commands in. The files do need those extensions though when you save them. Also note that while those extensions must be there, you can call the files anything you want (they don't need to be bilogex and parsex).
WARNING: When you install both PARSCALE and BILOG-MG on your machine, you need to make sure that copies of two files can be found in the directory C:/SSIDIR. Those two files are REGION.PSL and REGION.BLM. You should be able to find those two files in BILOGMG and PSL3 directories if needed. If these two files are not in the C:/SSIDIR directory then one or both of the programs will not run.
WARNING 2: Make sure the command files both end in .BLM or .PSL and not in .BLM.txt or .PSL.txt. You can see if they have the wrong name by going to the directory and typing DIR. If they do, you could rename them by using one of the following two lines:
rename bilogex.BLM.txt bilogex.BLM rename parscale.PSL.txt parscale.PSLWhen using notepad to save the files you edit, you will need to select "All Files" for the type instead of .txt.
There are many options for the command file, and some of them they will be progressively added throughout the semester.
To change the number of items in the exam, you must adjust the .BLM file in two places. They are indicated by being bold faced below. Simply replace the number 27s with the number of items you are working with.
EXAMPLE 1: Fitting the 3PL Model with Bilog >COMMENT Estimating a 3PL model with all the "default" settings. Note that you will have to change the directory names, and perhaps file names given below. >GLOBAL DFNAME='D:/778/SIMUL.DAT',NPARM=3,SAVE; >SAVE PARM='D:/778/BSIMUL.PAR',SCORE='D:/778/BSIMUL.SCO'; >LENGTH NITEMS=27; >INPUT TYPE=1,NIDCHAR=6; >ITEMS ; >TEST ; (6A1,27A1) >CALIB ; >SCORE ;
To change from estimating the 3PL model to estimating the Rasch model change the NPARM=3 on the GLOBAL line to NPARM=1.
BILOG-MG will produce three output files, plus any you requested in the command file. The names of these three files will begin with the name the command file and end in either .PH1, .PH2, or .PH3. Thus, if you use the bilogex sample command file you will get the five files: BILOGEX.PH1, BILOGEX.PH2, BILOGEX.PH3, BSIUML.PAR, and BSIMUL.SCO.
.PH1 gives which commands were read in, and which were unacceptable if there was an error. It will also give you the classical item analysis parameters.
.PH2 again gives some of the specifications that were made in the command file. It then gives some information on what some of the initial values were in estimation, as well as the log-likelihoods from the various iterations. Following right after the line that begins SUBTEST MAINTEST; ITEM PARAMETERS you will find the item parameter estimates, their standard errors, and a statistic to see if the IRF seemed to fit or not. BILOG calles the a parameter the slope, the b parameter the threshold, and the c parameter the asymptote. The intercept is -ab, and the dispersion is 1/a. The CHISQ gives both an approximate chi-square index and the corresponding p-value (PROB) [higher is better].
.PH3 gives some initial information and then gives the table of examinees. It lists the examinees by ID (for the example they are of the form 00001, 00002, etc...) and gives the number correct, the percent correct, the estimated ability, and the standard error of the estimated ability. It also gives the MEANS AND STANDARD DEVIATIONS OF SCORE ESTIMATES.
The other two files, the ones we called BSIMUL.PAR and BSIMUL.SCO contain the item parameter estimates and ability estimates from .PH2 and .PH3. The reason you might want these is that they don't have a lot of the extra information and would be easy to read into SAS or R or a Fortran simulation.
UNDER CONSTRUCTION! (But thats ok for now because I have the one you need written up already.)
PARSCALE will produce four output files, plus any you requested in the command file. The names of these four files will begin with the name the command file and end in either .PH0, .PH1, .PH2, or .PH3. Thus, if you use the parsex sample command file you will get the five files: PARSEX.PH0, PARSEX.PH1, PARSEX.PH2, PARSEX.PH3, PSIUML.PAR, and PSIMUL.SCO.
.PH0 gives which commands were read in, and which were unacceptable if there was an error. It also gives some information on what some of the initial values were in estimation.
.PH1 gives you classic item statistics in two tables. The first of the two tables is very hard to read though. In the second notice that the p-values (Response Mean) seems to be between 1 and 2!!! This is because PARSCALE counts a wrong answer as 1 and a correct as 2... so just subtract 1 from this value. Also, notice that the "Pearson Polyserial" correlation is the Biserial WITHOUT dropping the item in question (without the correction).
.PH2 After giving some of the log-likelihoods from the estimation, this file gives the parameter estimates and there standard errors in an easy to read format. It then gives some extra information concerning the estimation.
.PH3 gives some initial information and then gives the gives a table containing the ability estimates and there standard errors on the far right. The examinees are listed by ID (for the example they are of the form 00001, 00002, etc...). Below that table it also gives the MEANS AND STANDARD DEVIATIONS OF ABILITY DISTRIBUTIONS.
The other two files, the ones we called PSIMUL.PAR and PSIMUL.SCO contain the item parameter estimates and ability estimates from .PH2 and .PH3. The reason you might want these is that they don't have a lot of the extra information and would be easy to read into SAS or R or a Fortran simulation.
GGUM2000 is a free software program for the estimation of unfolding IRT models that can be found at: http://www.education.umd.edu/Depts/EDMS/tutorials/FreeSoftware.html. Scroll down the page to the section marked GGUM2000 Information and follow the link there to download and install the software.
The files that this program will put several files in the GGUM2000 directory on your C drive. One of these is the manual GGUMDOC.WPD in word perfect format. You can open this manual using microsoft word. (It is rather long, so if you print it out you should probably choose the four pages per sheet option.) The file that you need to select to run the programs is called MMLGUM.EXE and the output will be saved in the folder TEMPFILE.
To analyze a data set it is probably easiest to just save the command file (for example http://www.stat.sc.edu/~habing/courses/data/cappars.txt) and the data set (for example http://www.stat.sc.edu/~habing/courses/data/cap.dat) in the GGUM2000 directory. (The questions corresponding to this data set can be found at http://www.stat.sc.edu/~habing/courses/data/capdesc.html.)
Once the file is saved there, simply run the MMLGUM.EXE program. You will hit [enter] three times at the prompts, hit Y and return when it asks if you have a command file, and finally you will simply enter the name of the command file (e.g. cappars.txt) and hit return. The program will then run and put many files with unintuitive names in the TEMPFILE directory.
The descriptions of what these files contain can be found on pages 28-30 of the manual. In particular, the file FT16F001 contains the estimated parameters for the model and the file FT17F001 contains the theta estimates. For example, part of FT16F001 might look like the following:
ITEM# 1 INITIAL= 1.74293518 DELTA= 1.20332902 DSTD= .24610498 ALPHA= .38371085 ASTD= .06544841 CORR= -.49842280 THRESHOLD# 1 INITIAL= .00000000 TAU= .00000000 TAUSTD= .00000000 THRESHOLD# 2 INITIAL= -2.44110571 TAU= -1.81095524 TAUSTD= .66489221 THRESHOLD# 3 INITIAL= -1.95288457 TAU= -2.21699597 TAUSTD= .61059742 THRESHOLD# 4 INITIAL= -1.46466343 TAU= -.30961007 TAUSTD= .59568141 THRESHOLD# 5 INITIAL= -.97644229 TAU= -1.31886130 TAUSTD= .61225601 THRESHOLD# 6 INITIAL= -.48822114 TAU= .99970384 TAUSTD= .69366302This would indicate that the estimated delta was 1.203 (people with moderate positive levels of theta tend to endorse the item more than those with lower or higher thetas), a discrimination parameter of 0.38371085 (fairly low when compared to some of the other items), and Tau parameters of -1.81, -2.21, -0.30, -1.31, and 0.99 (although these aren't as interpretable just by looking at them). Note that the correlation reported is NOT the one between the response and predicted response.
In most cases, help with the computers in LeConte (NOT the programming) can be gained by e-mailing help@stat.sc.edu. If you are having programming trouble, just e-mail me. If you are having computer trouble with a computer outside LeConte... sorry...
For the printers on the first and second floor, printer paper is available in the Stat Department office. For printers on the third floor, paper is available in the Math Department office.
If you are using a PC restarting the machine will fix many problems, but obviously don't try that if you have a file that won't save or the like. DO NOT TURN OFF THE UNIX MACHINES.
In LeConte, if the problem is an emergency requiring immediate attention
see Jason Dew in room 415.
If Jason
is not available and it is an emergency see Minna Moore in room
417.
Flagrantly non-emergency cases may result in
suspension of computer privileges.