An analysis of differential tests as a predictor for learner performance in first year biosciences.

Abstract

This research project aims to establish the predictive validity of a set of aptitude tests for the first year Biosciences programme at a merging South African University. The study aims to address the problem of selection and placement to higher education and also to suggest how the results of aptitude tests might be used to inform curriculum development at first year level. The Differential Aptitude Test (DAT-L) was conducted on the 2004 Biosciences student cohort (228 students). The instructions that were given to the students followed the instructions laid out in Owen and Vosloo (1999). Normally there are ten tests in this battery of tests, however only eight of the tests were given to the students. The tests included; Vocabulary, Verbal reasoning, Non-Verbal reasoning: Figures, Calculations, Reading Comprehension, Comparisons, Price Controlling and Memory. Exam results of the participating students were collected. These included; Final Biosciences101 mark, Class mark (Biosciences101), Practical exam mark (Biosciences101) and Theory exam mark (Biosciences101). The final marks for Physics (mixed modules), Mathematics (mixed modules) Computer Sciences (CSCI 103) and Chemistry (CHEM131) were also collected. The indicator for the concept of academic potential in the Biosciences was taken as the marks achieved during the course of the first semester. Biographical information from the Student Management System (SMS) was also recorded this included; Race, Gender, Home language and Matric points. These various categories were compared against the students' performance in DAT-L tests. The average for the stanine points for a student of university "quality' is seven, according to Owen et al. (2000), while the average of this cohort of students is 5.16. The correlation for the different DAT tests when compared with the final result for Biosciences shows that the best correlation is with test 1 at r = .47292 (Vocabulary) while the worst correlation is with test 6 at r=.24722 (Comparison). The sequence of correlation is from Vocabulary through Reading Comprehension, Verbal Reasoning, Calculation, Memory, Price Controlling, Non-Verbal Reasoning: Figures, and finally Comparisons. The correlation of the average of all tests and the final result is a reasonable r = .50396. The best correlation with DAT and other subjects is Computer Science with r = .41165, with a declining correlation between Physics and the mark of r = .34085 followed by extremely poor correlations for Chemistry and Mathematics of r = .20313 and r = .08700 respectively. The best correlation that was obtained during this research was with the correlation between matric points and the average of all the DAT tests. While the correlation with DAT and the Final mark for Biosciences is r = .50396 that for the matric points is r = .57150. Females attained a significantly higher average on the DAT tests but the difference between female and male on their final results in Biosciences is not significant. White students perform significantly better than the other race designations when it comes to the final result. Within the categories African, Indian and Coloured there is no significant difference. When it comes to the different DAT tests, Africans perform significantly worse than the other three groupings and the Indian grouping did significantly worse when compared to Whites. If we were to use the DAT as a means of selecting students it would be fairly effective for biology but not for the other basic science subjects. The fact that the test would have to be administered to all applicants applying to do first year in the Science and Agriculture Faculty makes the choice of this set of tests questionable. Should the Faculty require a means of determining election into biological subjects then the tests show great promise. The pursuit of greater fairness, validity and reliability in selection is an ongoing quest.