An important statistic in classical item analysis is the item proportion-correct (item-analysis p-value), the proportion of the sample succeeding on a dichotomous item. In Winsteps, this value is interpreted as the average of the responses to the item for both dichotomous and polytomous items.

 

This proportion-correct-value is not a hypothesis test probability-value (statistical p-value). For hypothesis tests of fit, use the ZSTD values, which are probability-values expressed as unit-normal deviates.

 

PVALUE = NO Do not report the item proportion-correct-value.

 

PVALUE=YES Report item proportion-correct-values in Tables 6, 10, etc. and IFILE= (over-riding the existing default setting.)

 

Example: To parallel a classical analysis of an MCQ data set, it is desired to report the raw scores (including extreme persons and items), the point-biserials and the proportion-correct-values or average ratings.

 TOTALSCORE= YES ; report original total raw scores

 PTBISERIAL = YES  ; report point-biserial, not point-measure correlations

 PVALUE = YES  ; report proportion-correct-values

 

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

|ENTRY   TOTAL                       |   INFIT  |  OUTFIT  |PTBSE|PROP-|                 |

|NUMBER  SCORE  COUNT  MEASURE  ERROR|MNSQ  ZSTD|MNSQ  ZSTD|CORR.|VALUE| ITEM            |

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

|     1     35     35   -6.59    1.85| MINIMUM ESTIMATED MEASURE | 1.00| 1= 1-4          |

|     4     32     35   -4.40     .81| .90    .0| .35    .8|  .48|  .91| 4= 1-3-4        |

|    12      6     35    2.24     .55|1.16    .6|1.06    .5|  .26|  .17| 12=1-3-2-4-3    |

|    18      0     35    6.13    1.84| MAXIMUM ESTIMATED MEASURE |  .00| 18=4-1-3-4-2-1-4|