Example 8: Items with various rating scale or partial credit models

A 4 item test in which each item has a 9-level scoring rubric. We suspect that there are not really 9 levels of competence. After several WINSTEPS analyses, we choose to recode the rating scale or partial credit items by collapsing categories in order to increase measurement effectiveness (separation) and increase parameter stability. An objection to collapsing is that it violates the Rasch model. This is only true if the uncollapsed data strictly accord with the model. In fact, the collapsed data may fit the Rasch model better than the uncollapsed. We have to compare the collapsed and uncollapsed analyses to decide.

; This file is EXAM8.TXT

TITLE="Success and Failure Items"        

NAME1=6                                

ITEM1=1                                

NI=4                        ; 4 items                

ISGROUPS=1234        ; one item per grouping, same as ISGROUPS=0

IREFER=ABCD                ; the 4 items are to be recoded differently. Item 1 is type "A", etc.                

 CODES=123456789        ; the codes in the data file        

IVALUEA=333456666        ; the recoding for A-type items in IREFER=, i.e., Item 1

IVALUEB=333455555                

IVALUEC=333444444                

IVALUED=444456777                

&END

Maze

Passengers

Blocks

Egg race

END NAMES

5536 M Richard

4545 F Tracie

4345 M Walter

3454 M Blaise

4435 M Ron

....

Table 14.3 shows the recoding:

        ITEMS CATEGORY/OPTION/Distractor FREQUENCIES:  ENTRY ORDER

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

|ENTRY   DATA  SCORE |     DATA   |     USED     AVERAGE OUTF|           |

|NUMBER  CODE  VALUE |  COUNT   % |  COUNT   %   MEASURE MNSQ|  ITEM     |

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

|    1   2         3 |      1   3 |      1   3     -2.39   .3|Maze       |

|        3         3 |      7  22 |      6  19     -1.57   .7|           |

|        4         4 |     10  31 |     10  32      -.54   .5|           |

|        5         5 |     10  31 |     10  32      1.23   .6|           |

|        6         6 |      3   9 |      3  10      2.42   .7|           |

|        7         6 |      1   3 |      1   3      2.78   .5|           |