The purpose of this study was to evaluate two linear equating methods where the abilities of nonrandom equating groups vary. The two equating methods investigated in the study were the Tucker and Levine methods for equally-reliable tests and internal anchor forms. The study was particularly relevant for minimum competency and basic skills testing programs since conditions were simulated where successive equating groups increased in mean ability while decreasing in variance. / Six equating groups of approximately 300 examinees each were selected according to scores on the communication section of the 1978 Florida Fifth-Grade State Student Assessment Test. The examinees' scores were selected using a stratified random sampling plan intended to yield six ability groups for scores on the mathematics section of the test. The mathematics score distributions of the groups varied from flatter-than-normal to a skew of -2.420. The average mathematics p-values ranged from .64 to .92. / The mathematics scores of the groups were equated to the three lowest-ability groups using each of the two methods in order to simulate conditions where successive equating groups increase in ability from a group whose scores are used to form the criterion scale. Since a test was equated to itself, the differences between the equated and observed scores of examinees was considered to be error produced by the equating methods. / Six anchor tests were selected to investigate the accuracy of equating with the two methods where the anchor form is not content--representative of the operational form. One anchor form was constructed to be parallel in content with the operational form. The five nonrepresentative forms varied in difficulty from easy to hard. / The Levine method generally produced smaller average, absolute errors of conversions of examinees scores than the Tucker method except where the abilities of the equating groups were similar. The Levine method also generally produced less error for the conversion of group means. However, the Levine method conversion lines were characterized by high, positive intercept values which generally increased as the abilities of the equating groups diverged. Therefore, low scores were equated with considerable error. / Except where the differences between the ability groups were greatest, both equating methods generally equated the scores in the upper 80% of the groups with error of less than one raw score point. Also, the distances between the converted means and the actual raw score means of the equated groups generally did not vary by more than 5%. / In light of these data, it was concluded that the use of either equating method would probably not result in serious error for comparisons of equating group means where the abilities of equating groups vary substantially. Furthermore, either method could be used if critical decision points occur high on the score scale. However, the Levine method would be recommended in each of these cases since it consistently produced less average error in this study. Caution would be advised for the use of either method where critical decision points occur low on the scale. / In spite of the fact that three of the nonrepresentative anchor forms correlated better than .95 with the operational form, each nonrepresentative anchor form produced a substantially larger average error than the representative anchor form for each method. These results demonstrated the necessity of constructing anchor forms which are content-representative of the operational forms. / Source: Dissertation Abstracts International, Volume: 41-11, Section: A, page: 4686. / Thesis (Ph.D.)--The Florida State University, 1980.
| Identifer | oai:union.ndltd.org:fsu.edu/oai:fsu.digital.flvc.org:fsu_74347 |
| Contributors | GREEN, JOHN COLLINS., Florida State University |
| Source Sets | Florida State University |
| Detected Language | English |
| Type | Text |
| Format | 140 p. |
| Rights | On campus use only. |
| Relation | Dissertation Abstracts International |
Page generated in 0.0022 seconds