Separate versus Concurrent Calibration Methods in Vertical Scaling.
Saved in:
| Title: | Separate versus Concurrent Calibration Methods in Vertical Scaling. |
|---|---|
| Language: | English |
| Authors: | Karkee, Thakur, Lewis, Daniel M., Hoskens, Machteld, Yao, Lihua, Haug, Carolyn |
| Peer Reviewed: | N |
| Page Count: | 47 |
| Publication Date: | 2003 |
| Document Type: | Reports - Research Speeches/Meeting Papers |
| Descriptors: | Estimation (Mathematics), Mathematics Tests, Scaling, Scoring, State Programs, Testing Programs |
| Abstract: | Two methods to establish a common scale across grades within a content area using a common item design (separate and concurrent) have previously been studied under simulated conditions. Separate estimation is accomplished through separate calibration and grade-by-grade chained linking. Concurrent calibration established the vertical scale in a single step by simultaneously estimating parameters for all items at all grades. These methods, and a third hybrid method, pairwise concurrent estimation, were examined in this study, using operational test data. Data were obtained from responses to the 2002 Colorado Student Assessment Program (CSAP) mathematics assessments in grades 5 through 10. These assessments had been constructed to have at least 20 items in common between adjacent grades. The data set for each grade consisted of 10,000 cases, randomly selected from the population of approximately 54,000 students with valid scores at each grade. The 2002 CSAP mathematics scales were placed on a common scale spanning grades 5 through 10 using the three methods. Standard analyses of calibration output indicated that the separate estimation method produced consistently better results than did the concurrent of pairwise concurrent estimation. (Contains 9 figures and 19 references.) (SLD) |
| Entry Date: | 2004 |
| Accession Number: | ED478167 |
| Database: | ERIC |
| Abstract: | Two methods to establish a common scale across grades within a content area using a common item design (separate and concurrent) have previously been studied under simulated conditions. Separate estimation is accomplished through separate calibration and grade-by-grade chained linking. Concurrent calibration established the vertical scale in a single step by simultaneously estimating parameters for all items at all grades. These methods, and a third hybrid method, pairwise concurrent estimation, were examined in this study, using operational test data. Data were obtained from responses to the 2002 Colorado Student Assessment Program (CSAP) mathematics assessments in grades 5 through 10. These assessments had been constructed to have at least 20 items in common between adjacent grades. The data set for each grade consisted of 10,000 cases, randomly selected from the population of approximately 54,000 students with valid scores at each grade. The 2002 CSAP mathematics scales were placed on a common scale spanning grades 5 through 10 using the three methods. Standard analyses of calibration output indicated that the separate estimation method produced consistently better results than did the concurrent of pairwise concurrent estimation. (Contains 9 figures and 19 references.) (SLD) |
|---|
