Trang 1
A Pilot of Aptitude Testing for University Entrance Angus S. McDonald Paul E. Newton Chris Whetton with an appendix by Samantha E. Higgs
Nội dung tài liệu
Trang 2
Contents List of tables List of figures Executive summary 6 Background 8 Introduction 8 Goals of the present study 11 Methodology 13 Samples 13 Materials 13 Results 15 Response rates 15 Treatment of data 15 Descriptive statistics 17 Multilevel modelling - overview 25 Multilevel modelling - findings 26 Analysis of SAT functioning 36 Conclusions 41 References 44 Appendix 1: Score distributions of main study variables 45 Appendix 2: Classification of universities and colleges 49 Appendix 3: Details of multilevel modelling 50 Appendix 4: Item functioning data (IRT analysis) 70 Appendix 5: Item functioning data (classical test analysis) 74
Trang 3
List of Tables Table 1: Number of schools returning completed materials, and number 15 of students meeting minimum data requirements Table 2: GCSE, A-level and SAT I: Reasoning Test scores by sample 17 Table 3: GCSE, A-level and SAT I: Reasoning Test scores by sex 18 Table 4: GCSE, A-level and SAT I: Reasoning Test scores by ethnicity 19 Table 5: GCSE, A-level and SAT I: Reasoning Test scores by parental 20 socio-economic status Table 6: GCSE, A-level and SAT I: Reasoning Test scores by intentions 21 at the end of current course Table 7: GCSE, A-level and SAT I: Reasoning Test scores by 23 classification of universities and colleges Table 8: Correlations of GCSE, predicted and attained A-levels, SAT I: 25 Reasoning Test scores and socio-economic status by sample Table 9: SAT I: Reasoning Test scores by GCSE grade bands and 28 sample Table 10: SAT I: Reasoning Test scores by A-level grade bands and 29 sample Table 11: Percentage of students above selection thresholds 31 Table 12: Cronbach's alpha reliabilities for the SAT I: Reasoning Test 37 Table 13: Correlations between IRT parameters in British and American 38 students
Trang 4
List of Figures Figure 1: Regression of A-level on SAT I: Reasoning Test and SAT I: 32 Reasoning Test on A-Level Figure 2: Regression of A-level on SAT I: Reasoning Test for 33 independent schools Figure 3: Regression of A-level on SAT I: Reasoning Test for low- 34 achieving schools Figure 4: Regression of A-level on SAT I: Reasoning Test for high- 35 achieving schools Appendices Figure 3.1 Normalised coefficients 53 Figure 3.2 Random variances in A-levels at different levels with and 53 without background variables Figure 3.3 Normalised coefficients when fitting SES as continuous 56 Figure 3.4 Random variances, SES as continuous 56 Figure 3.5 Normalised coefficients, fitting four separate models 57 Figure 3.6 Normalised coefficients fitting math score 58 Figure 3.7 Random variances fitting math score 59 Figure 3.8 Normalised coefficients fitting verbal score 59 Figure 3.9 Random variances fitting verbal score 60 Figure 3.10 Normalised coefficients fitting a three-level model 61 Figure 3.11 Random variances fitting a three-level model 62 Figure 3.12 Scatterplot of A-level v. SAT I: Reasoning Test for all 63 samples Figure 3.13 Scatterplot of GCSE v. SAT I: Reasoning Test for all 64 samples Figure 4.1 Scatterplot of verbal IRT difficulties for British and 72 American students Figure 4.2 Scatterplot of IRT math difficulties for British and 73 American students
Trang 5
Acknowledgements The project team for this work consisted of: Design and secretarial support: Liz Gibson and Jackie Hill Administration of project materials: Keren Beddow, Maria Charles, and John Hanson Statistical analysis: Samantha Higgs, Rachel Felgate, Ian Schagen, Dougal Hutchison and Sally Green The NFER also gratefully acknowledges the assistance of Educational Testing Services for providing and scoring the SAT I: Reasoning Test. Note: The SAT I: Reasoning Test has two elements, a verbal section and a 'math' section. Since this American English term is that used in the test itself, it has been referred to in this way throughout this report.
Trang 6
p. 6 Executive summary This report presents the findings from a study examining the association between the Scholastic Assessment Test (SAT) and A-level attainment. This research was commissioned by The Sutton Trust and conducted by the National Foundation for Educational Research (NFER). Recent media debate on the British university admissions system has highlighted the American SAT as a potential way of identifying able students from less privileged backgrounds. In doing so, it was offered as one way of redressing the imbalance in students from state and independent schools in some of the highest-ranked universities. However, as no recent British studies have been conducted on the SAT, there was a lack of empirical evidence for many of the claims that were made for it. The present study was undertaken to partially redress this, by looking at the association between the SAT and A-level grades, and whether this association varied according to background factors. Three samples of students participated in this study: high- and low-achieving schools, sampled on the basis of their GCSE results, and a sample of selective independent schools provided by The Sutton Trust. Participating students completed a questionnaire which collected personal details, their GCSE and predicted A-level grades, and their intended destination when they had finished their current course. Students also completed a short version of the SAT known as the SAT I: Reasoning Test, which contained verbal and math sections. Attained A-level results were collected when available. Completed materials were received from 1,295 students. The effects of a range of background factors on GCSE and A-level grades and the SAT I : Reasoning Test scores were studied. These included sample (high-attaining, low- attaining and independent schools), sex, parental socio-economic status, intended destination and students’ intended university or college if they planned to go on to higher education. The correlations between study variables were also examined. The main tests of the study hypotheses were conducted with multilevel modelling, with mean attained A-level grade as the dependent variable. This revealed that the SAT I: Reasoning Test was modestly associated with A-level grades, but there was no evidence that the association differed according to background factors such as ethnicity, parental socio- economic status and sample. The SAT may be of value in predicting university performance but the data provided no evidence that it was able to assess potential for study at higher education, independently of a student’s social and educational experiences.
Trang 7
p. 7 The modest correlation between the SAT I: Reasoning Test and A-level grades meant that if the SAT score were used for selection, roughly the same proportion of students would be considered from each type of school as with A-Levels. However, these might not be the same individuals. Using the SAT scores in addition to A-levels would increase the number of students selected for all types of school, but the percentage increase would be greatest for students from low-attaining schools. Further analyses explored the functioning of the SAT I: Reasoning Test in British students. These showed that the SAT I: Reasoning Test provided a coherent assessment of verbal and math reasoning ability, and that individual items appeared to function similarly in British and American students. Item-level analyses found little evidence of bias in SAT items between males and females, or the three samples of schools. Although this study failed to find evidence that background factors differentially affected A- level grades and SAT I: Reasoning Test scores, SAT scores were only modestly associated with A-level grades. This indicates that the SAT I: Reasoning Test assesses a distinct construct from A-levels, and that further investigation of aptitude testing as a predictor of university performance is warranted.
Hiển thị 7 trang đầu. Tải tài liệu để xem đầy đủ.
Bình luận (0)
Chưa có bình luận nào. Hãy là người đầu tiên!