Predicting success of a beginning computer course using logistic regression (abstract only)

Abstract

Enrollment in computer science classes have grown at an alarming rate despite often inadequate university resources and a shortage of trained computer science faculty. While the number of computer science faculty. While the number of computer science majors may now be leveling off or even declining, many other disciplines are requiring students to possess computer skills. An all too high percentage of students do not possess a sufficient academic background which thereby causes a high attrition rate in computing courses. These failures represent a loss of both university and student resources.Whether choosing a selected group of students from a large number trying to enroll in a specific course or program or attempting to offer different prerequisite computer classes based on student need and aptitude, some method for predicting performance is required. This study was conducted to see what variables might be predictors of success in a first computer science course.In this study, success in the beginning course for computer science majors and minors was defined as earning a grade of C or better, while failure was considered as receiving a D, an F, or withdrawing from the course. The authors believe, from their experience, that what one instructor would evaluate as A caliber work another instructor might rate only as B performance. However, instructor agreement improves when performance is rated as acceptable (C or above) or not acceptable (D, F or withdrawal). The predictors used included success in a beginning calculus course, SAT quantitative and verbal scores, high school class percentile rank, sex of the student, quarter in which the course was taken, class status of the student, and the instructor who taught the course. The sample was 262 students enrolled in different sections of the beginning course during three different quarters 1984-85.The stepwise logistic regression model was chosen because the dependent variable was dichotomized as described above and the predictor variables were either interval or categorical. High school class percentile rank was computed by subtracting the rank of the student in his or her graduating class from the size of the graduating class, dividing this difference by the size of graduating class. This variable, then, is interval, SAT verbal and SAT quantitative scores are interval variables. Success in the beginning calculus course was defined the same way as success in the beginning computer science course was defined, with a third category added for students who had not attempted calculus. Thus, success in the beginning calculus course, sex of the student, the quarter in which the course was taken, and the instructor are categorical variables. SAT scores, the mathematics prerequisite, and sex of the student have been tested by others as predictors for success in computer science. [1, 2, 3, 4, 5].The other studies noted above assume that course grades are on an equal interval scale which allowed the use of an ordinary regression model. The predictor variables in an ordinary regression model must either be continuous or have only two categories. The model used in the present study allows the stepwise technique utilized in the studies cited above for ordering predictor variables into the equation, but also allows categorical predictors to have more than two states if they are not independent.The results of this study indicate that the best predictor of success in the computer science course is success in a calculus course. This result supports the findings in other studies [2, 5]. The second best predictor found was high school class percentile rank. The only other variables that could account for variation in the dependent variable were sex of the student and class status of the student. A prior study also suggests that male students performed better in computer science courses. [2]The authors have concluded from the results that the reasoning skills necessary for success in calculus are also important for success in computer science. Another trait for success is the competitiveness of the student. This trait is measured in some degree by the high school percentile rank.