A study of the role that specification error plays in the ranking of meritorious schools produced by the "merit as residual" and Analysis of Covariance (ANCOVA) regression-based procedures was conducted. Two specification issues, inclusion of relevant variables and functional form, were examined in the context of state and district level applications of merit assessment. / For the state level analysis, 100 schools were selected from the High School and Beyond data set to reflect a population of high schools in a state level setting. To examine what effect the removal of relevant variables has on the ranking of meritorious schools, various degrees of misspecification were introduced into a final analysis model. This final model was assumed to reflect the "true" ranking of schools. Misspecification was introduced by removing variables, or blocks of variables, from the model. School residuals produced by the misspecified models were compared with the school residuals produced by the "true" model to determine whether changes in school rankings occurred. / Six of the 100 schools used in the state level analysis were selected to reflect a population of high schools in a district level setting. An ANCOVA model was developed. This model provided a set of adjusted means which were used to rank the six schools. To examine misspecification attributed to functional form, the ANCOVA model was extended to an Aptitude-Treatment Interaction (ATI) model, which contained interaction terms. This ATI model, which was assumed to reflect the "true" model of student achievement, was used to provide three school rankings for different subgroups of students. / This study concluded that school rankings are dependent on the specification of the model used as the basis for making the merit assessment. The findings provided evidence that important ranking changes occurred even when relatively small increments in the amount of variance explained were associated with a block or individual variable. This conclusion, along with the current state of model development, suggests that school rankings produced by these procedures may not provide a close approximation of reality. / Source: Dissertation Abstracts International, Volume: 49-07, Section: A, page: 1755. / Major Professor: Richard Tate. / Thesis (Ph.D.)--The Florida State University, 1988.
| Identifer | oai:union.ndltd.org:fsu.edu/oai:fsu.digital.flvc.org:fsu_77805 |
| Contributors | Douglas, Kathy Aileen., Florida State University |
| Source Sets | Florida State University |
| Language | English |
| Detected Language | English |
| Type | Text |
| Format | 309 p. |
| Rights | On campus use only. |
| Relation | Dissertation Abstracts International |
Page generated in 0.0022 seconds