The IRT Procedure

Example 65.5 Subject Scoring

This example also uses the American Quality of Life Survey data introduced in Example 65.1. The purpose of this example is to show you how to score subjects. You can score subjects in two different ways by using the IRT procedure. If you want to fit the model and score the subjects simultaneously, you simply use the OUT= option in the PROC IRT statement. The following statements fit an IRT model to the data set IrtUni and then score all the subjects in that data set based on the parameter estimates. Factor scores along with the original data in IrtUni are saved to a SAS data set called IrtScore. The first five observations in the OUT= data set are displayed in Output 65.5.1. 

proc irt data=IrtUni out=IrtScore;
   var item1-item8;
run;

proc print data=IrtScore(obs=5);
run;

Output 65.5.1: Factor Scores for the First Five Observations

Obs item1 item2 item3 item4 item5 item6 item7 item8 _Factor1
1 1 0 0 0 1 1 2 1 -0.38031
2 1 1 1 1 1 3 3 3 1.69165
3 0 1 0 0 1 1 1 1 -0.72590
4 1 0 0 1 0 1 2 3 0.12298
5 0 0 0 0 0 1 1 1 -1.59052 




            

            


            
In applied research, it is not uncommon to want to save the parameter estimates (or item calibration result) and use them
               later to score new subjects without refitting the model. To accomplish this task, first you use the OUTMODEL=
                option in the  PROC IRT
                statement to save the parameter estimates as follows: 
            


proc irt data=CalData outmodel=IrtModel;
   var item1-item8;
run;



 In the preceding statements, the data set CalData contains 500 random samples from the original data set IrtUni. 
            

            
Then you use the following statements to score new subjects in the data set that is specified in the DATA=
                option, where the data set NewSub contains 50 random samples from the IrtUni data set. In the following statements, you use the INMODEL=
                option in the  PROC IRT
                statement to input the model specification and parameter estimates from a previous analysis. To score the subjects without
               refitting the model, you also need to specify the SCORE suboption. Output 65.5.2 contains the first five observations in the OUT= data set. 
            


proc irt data=NewSub inmodel(score)=IrtModel out=IrtScore2;
run;

proc print data=IrtScore2 (obs=5);
run;



  

            

               
Output 65.5.2: Factor Scores for the First Five Observations

               

                  

                     
                     

                        
                        
                        

                           
                           

                              
                              
















Obs
                                       
                                       Replicate
                                       
                                       item1
                                       
                                       item2
                                       
                                       item3
                                       
                                       item4
                                       
                                       item5
                                       
                                       item6
                                       
                                       item7
                                       
                                       item8
                                       
                                       _Factor1
                                       
                                    




1
                                       
                                       1
                                       
                                       1
                                       
                                       1
                                       
                                       1
                                       
                                       1
                                       
                                       1
                                       
                                       2
                                       
                                       2
                                       
                                       3
                                       
                                       1.14800
                                       
                                    



2
                                       
                                       1
                                       
                                       1
                                       
                                       0
                                       
                                       1
                                       
                                       1
                                       
                                       1
                                       
                                       1
                                       
                                       3
                                       
                                       1
                                       
                                       0.45932
                                       
                                    



3
                                       
                                       1
                                       
                                       1
                                       
                                       1
                                       
                                       1
                                       
                                       1
                                       
                                       1
                                       
                                       2
                                       
                                       3
                                       
                                       2
                                       
                                       1.11391
                                       
                                    



4
                                       
                                       1
                                       
                                       1
                                       
                                       1
                                       
                                       1
                                       
                                       1
                                       
                                       1
                                       
                                       1
                                       
                                       1
                                       
                                       2
                                       
                                       0.56333
                                       
                                    



5
                                       
                                       1
                                       
                                       0
                                       
                                       0
                                       
                                       0
                                       
                                       0
                                       
                                       1
                                       
                                       2
                                       
                                       2
                                       
                                       3
                                       
                                       -0.20233
                                       
                                    






                           
                        

                        



                     
                     
                  





            

            


            
If you do not specify the SCORE suboption, PROC IRT uses the parameter values specified in the INMODEL=
                option as initial values, refits the model, and then uses the new parameter estimates to score the subjects.