Direct Writing Prediction Models Identify At-Risk Writers
Saved in:
| Title: | Direct Writing Prediction Models Identify At-Risk Writers |
|---|---|
| Language: | English |
| Authors: | Conrad, Charles James Harding, II |
| Source: | THAITESOL Journal. Jan-Jun 2020 33(1):57-71. |
| Availability: | Thailand TESOL Organization. Language Institute Building, Thammasat University, 2 Prachan Road, Pranakhorn, Bangkok, Thailand 10200. e-mail: journal.thaitesol@gmail.com; Web site: https://www.tci-thaijo.org/index.php/thaitesoljournal/index |
| Peer Reviewed: | Y |
| Page Count: | 15 |
| Publication Date: | 2020 |
| Document Type: | Journal Articles Reports - Research Tests/Questionnaires |
| Education Level: | Elementary Education Secondary Education |
| Descriptors: | Foreign Countries, Elementary School Students, Secondary School Students, Writing (Composition), Writing Evaluation, Scoring Rubrics, Writing Skills, Prediction, At Risk Students, Writing Tests, Models, Scores, Longitudinal Studies |
| Geographic Terms: | Thailand |
| ISSN: | 2286-8909 |
| Abstract: | Developing a sufficient level of writing proficiency takes time. It is also a complex skill difficult to measure. The history of writing assessments reveals changing views of construct validity, reliability and interpretation of results. This study used a binary logistic regression model with seven years of grades 3 to 12 annual direct writing assessments scored with the Oregon six traits rubric from 2012-2018. Three predictive models were developed to show how likely it would be for a participant to reach writing proficiency, and how long it may take to meet that expectation. The research question was, "To what extent can the Annual Writing Assessment scored with the six-traits writing rubric identify at-risk writers from Grades 3-12 at the International Community School Bang Na, Thailand?" Results indicated the bio-data did not prove significant in any of the three models. Increased direct writing data input improved the prediction accuracy. In the Year 1 model, only the average test scored proved significant. In the Year 2 model, the trait of conventions proved significant as one of the independent variables along with the first- and second-year averaged test scores, and the difference between those averages. In the Year 3 model, conventions and sentence fluency proved significant along with the first- and third-year averaged test scores. The process of developing the predictive models, and the results for identifying at-risk writers are presented in this quantitative longitudinal research. |
| Abstractor: | As Provided |
| Entry Date: | 2020 |
| Accession Number: | EJ1257618 |
| Database: | ERIC |
| Abstract: | Developing a sufficient level of writing proficiency takes time. It is also a complex skill difficult to measure. The history of writing assessments reveals changing views of construct validity, reliability and interpretation of results. This study used a binary logistic regression model with seven years of grades 3 to 12 annual direct writing assessments scored with the Oregon six traits rubric from 2012-2018. Three predictive models were developed to show how likely it would be for a participant to reach writing proficiency, and how long it may take to meet that expectation. The research question was, "To what extent can the Annual Writing Assessment scored with the six-traits writing rubric identify at-risk writers from Grades 3-12 at the International Community School Bang Na, Thailand?" Results indicated the bio-data did not prove significant in any of the three models. Increased direct writing data input improved the prediction accuracy. In the Year 1 model, only the average test scored proved significant. In the Year 2 model, the trait of conventions proved significant as one of the independent variables along with the first- and second-year averaged test scores, and the difference between those averages. In the Year 3 model, conventions and sentence fluency proved significant along with the first- and third-year averaged test scores. The process of developing the predictive models, and the results for identifying at-risk writers are presented in this quantitative longitudinal research. |
|---|---|
| ISSN: | 2286-8909 |
