The Effects of Multidimensional Polytomous Response Data on Unidimensional Many-FACET Rasch Model Parameter Estimates.
Saved in:
| Title: | The Effects of Multidimensional Polytomous Response Data on Unidimensional Many-FACET Rasch Model Parameter Estimates. |
|---|---|
| Language: | English |
| Authors: | Wang, Shudong, Wang, Ning |
| Peer Reviewed: | N |
| Page Count: | 24 |
| Publication Date: | 2003 |
| Document Type: | Reports - Research Speeches/Meeting Papers |
| Descriptors: | Estimation (Mathematics), Item Response Theory, Simulation |
| Abstract: | When categorical responses were simulated from a Multidimensional Many-FACETS Rasch Compensatory Model (MMFRCM), the effects of ability, task difficulty, and step difficulty estimates with the unidimensional Many-FACETS Rasch Model (MFRM; Linacre, 1999) were examined in terms of three error indexes, average absolute difference (AAD), bias, and root mean square error (RMSE). The results show that violating unidimensional assumptions does have an effect on parameter estimation. However, the degree to which estimation shows robustness or not varies dramatically. The conclusion is that the complex nature of the model and data must be clearly understood to determine under which conditions the model should be applied and how well the parameters associated with the model can be estimated reliably. This study provides strong evidence that indicates the nature of MFRM performance when model assumption is violated.(Contains 11 tables and 44 references.) (Author/SLD) |
| Entry Date: | 2004 |
| Accession Number: | ED478079 |
| Database: | ERIC |
| Abstract: | When categorical responses were simulated from a Multidimensional Many-FACETS Rasch Compensatory Model (MMFRCM), the effects of ability, task difficulty, and step difficulty estimates with the unidimensional Many-FACETS Rasch Model (MFRM; Linacre, 1999) were examined in terms of three error indexes, average absolute difference (AAD), bias, and root mean square error (RMSE). The results show that violating unidimensional assumptions does have an effect on parameter estimation. However, the degree to which estimation shows robustness or not varies dramatically. The conclusion is that the complex nature of the model and data must be clearly understood to determine under which conditions the model should be applied and how well the parameters associated with the model can be estimated reliably. This study provides strong evidence that indicates the nature of MFRM performance when model assumption is violated.(Contains 11 tables and 44 references.) (Author/SLD) |
|---|
