A Longitudinal Evaluation of a Best Practices CS1

Abstract

Over a decade ago, the CS1 course for students without prior programming experience at a large research-intensive university was redesigned to incorporate three best practices in teaching programming: Media Computation, Pair Programming, and Peer Instruction. The purpose of this revision was to improve the quality of the course, appeal to a larger student body, and improve retention in the major. An initial analysis of the course indicated an increase in pass rates and 1-yr retention of students in the major. Now that time has passed and those students impacted by the revision have had time to graduate, this longitudinal study revisits and expands on these prior findings through examining student outcomes over a twelve year period (2001 through 2013). The student outcomes examined include failure rates in CS1, retention rates in the major, rates of switching into the major, time to degree, and performance in subsequent major courses. We compare these findings against similar metrics collected for another CS1 course at the same institution that caters to students with prior programming experience and did not make changes during this same time period. Overall, the inclusion of media computation, pair programming, and peer instruction corresponds to a significant improvement in passing rates for CS1 as well as retention of majors from CS1 through graduation. In turn, there is no indication that this larger group of students experienced any harm in terms of lower grades in upper-division courses or their time to degree.