NAEP Technical DocumentationNAEP Assessment IRT Parameters

			Arts Assessment IRT Parameters Civics Assessment IRT Parameters Economics Assessment IRT Parameters Geography Assessment IRT Parameters Mathematics Assessment IRT Parameters Reading Assessment IRT Parameters Science Assessment IRT Parameters Technology and Engineering Literacy Assessment IRT Parameters U.S. History Assessment IRT Parameters Writing Assessment IRT Parameters Mathematics Long-Term Trend Assessment IRT Parameters Reading Long-Term Trend Assessment IRT Parameters

During scaling, Item Response Theory (IRT) parameters are estimated using data from the current assessment and the most recent past assessment of the same subject, if that past assessment was developed according to the same assessment framework. For items fitting the two-parameter IRT model, "a" and "b" parameters are estimated. For items fitting the three-parameter model, "a," "b," and "c" are estimated. For the generalized partial-credit model, "a," "b," and "d" parameters are estimated. There is an acceptable range of values for each IRT parameter. The "a" parameter typically ranges from 0 to 2. Generally, this value should never be negative. The "b" parameter typically varies from -3 to 3, and the "c" parameter varies from 0 to 1. The range of values for the "d" parameter is generally between -3 and 3, but can at times vary from -8 to 8. Items that are functioning poorly on a scale are identified early (i.e., during pilot tests and item analysis) and dropped from that scale. As with other IRT scaling procedures, person parameters are also estimated while the items are scaled; however, NAEP does not make use of these estimates, because group results based directly on these individual student parameters are inconsistent (Mislevy 1991). Note that the item parameters here are provided in the metrics used for the original calibration of the scales.

Printer-friendly Version