Formative assessment by first-year chemistry students as predictor of success in summative assessment at a South African university

Abstract

This study investigated whether formative assessment is a predictor of summative assessment in a university first-year chemistry class. The sample comprised a total of 1687 first-year chemistry students chosen from the 2011 and 2012 cohorts. Both simple and multiple linear regression (SLR and MLR) techniques were applied to perform the primary aim of the research. In order to apply the above mentioned techniques, a selection criterion was executed on the 1687 sample, after which 1519 cases remained for the analysis. The study revealed a statistically significant SLR model, suggesting that formative assessment (FA) can, at 45.4% (that is R² = 0.454) level of accuracy, predict the summative assessment (SA) of students in a university first-year chemistry class. Furthermore, the results of two MLR models discovered that SA can be predicted by using theory marks at 57.1% (that is R² = 0.571) level of accuracy, and average semester test marks at 59.4% (that is R² = 0.594) level of accuracy. The aforementioned domino effects suggest that the semester tests marks are more efficient, among other marks, in predicting the SA marks of students in a chemistry department at a South African traditional university; accordingly recommends that more effort be made in preparing students for their semester marks. In addition the study found that of the 1519 students who had the 40% subminimum entry requirement for summative assessment, 765 (51.4%) passed the summative assessment, that of the 277 whose formative assessment mark was between 40% and 49% (both inclusive), 6 (2.2%) passed the summative assessment, that of the 1208 whose formative assessment mark was at least 50%, 725 (60%) passed the summative assessment and, finally, that of the 34 students whose formative assessment mark was at least 75% only 8 were able to retain their grades. Overall, linear regression and probability analyses suggest that FA does predict the SA mark, but it is best when the AVRSEMT mark is used.