Running head: MATH LEVELS AND COMPUTER SCIENCE Math Levels of Computer Science Students Rick Fillman Institutional Research Analyst Planning and Research Office June 2012 MATH LEVELS OF COMPUTER SCIENCE Page 2 of 8 Introduction The Computer Science (CS) faculty is considering a change to the course prerequisites for CS-11 (Introduction to Programming Concepts and Methodology, C++) and CS-12J (Introduction to Programming Concepts and Methodology, Java). At present the recommended preparation for both courses lists “Eligibility for ENGL 100 and READ 100” and “Recommended Preparation: CS-1 and MATH-154” (Elementary Algebra). The faculty are considering changing the current recommended prep to a prerequisite of Math-152 (Intermediate Algebra). Note that a prerequisite of MATH-152 would require that students complete, or demonstrate mastery of, both Elementary and Intermediate Algebra prior to enrolling in these CS courses. If such a prerequisite were instituted, how many students might be impacted? Would course success rates increase? Methodology Student records for eight years of enrollments in these Computer Science courses were gathered. Students’ math levels at the time of the student’s first enrollment1 in either CS-11 or CS-12J were determined. These students become the focus of this research. Math levels are determined by assessment/placement score, satisfactory completion of an equivalent course elsewhere, or by satisfactory completion of a course in the Cabrillo mathematics sequence. Data from assessment/placement scores and equivalencies from other institutions were gathered from the Datatel student information system. Data for Cabrillo math enrollments where the student successfully completed a Math course (grade of “A”, ”B”, “C” or P”) are assembled from the Cabrillo research data warehouse. These data are chronologically aligned with the CS course enrollments, and the student’s math level at the time of the CS enrollment is determined by the most recently completed Math course prior to the term of the CS enrollment. In cases where there is no satisfactory math completion record (the student didn’t take math at Cabrillo, or did and received a grade of D, F, or W), then the students’ assessment/placement and/or equivalency level, if available, is used to 1 Repeat enrollments in CS courses are excluded from the analysis. The repeat rate for the time period studied is about 8.9% for CS -11, and about 5.4% for CS-12J. Page 2 MATH LEVELS OF COMPUTER SCIENCE Page 3 of 8 determine the math level. For this study, the math level of some students could not be determined. This occurs in cases where a student never attempts to enroll in a math course at the college, or for some reason never undergoes assessment/placement in math. Under the proposed prerequisite, these students would be required to undergo assessment, so their status could be determined. How would enrollment in CS courses be affected? If Math-152 Intermediate Algebra was established as a prerequisite for these CS courses, about 80% of first-time enrollees both CS-11 and CS-12J would clearly have met the hypothetical prerequisite by the time of their enrollment in CS. This leaves possibly one in five students whose course enrollment might be precluded or delayed, if such a prerequisite were established. The chart below shows the number of students who would have met or not met this hypothetical prerequisite at the time of their first enrollment in CS-11 or CS-12J, respectively. These data are compiled from eight years of enrollment records from academic years 2003-04 through 2010-11. Historical percentage of students meeting or not meeting a hypothetical Intermediate Algebra prerequisite for CS Headcount Percent CS-11 Prereq. met 749 80.2% Prereq. not met 78 8.4% Prereq. status unknown 107 11.5% Headcount Percent CS12J Prereq. met 744 79.6% Prereq. not met 76 8.1% Prereq. status unknown 115 12.3% Disproportionate Impact When considering ethnicity and gender balance in these courses, it is helpful to review the extent to which the profile of CS students diverges from the profile of the overall college population. Although Latinos constitute in excess of 30% of the Cabrillo College student body and females constitute more than 52% of students, participation Page 3 MATH LEVELS OF COMPUTER SCIENCE Page 4 of 8 rates for these groups in CS courses is much lower. The chart below shows participation of these groups in the two CS courses studied. Actual partipation rates in CS Courses CS-11 18% 18% Latino Female CS12-J 13% 21% The introduction of the hypothetical prerequisite would likely not alter these proportions. In other words, when the ethnic and gender breakout of students who meet or do not meet the hypothetical are examined, though differences among the groups is observed, the differences are not statistically significant. The potential for disproportionate impact appears small. Percentages of students by under-represented status and by gender who would have met/not met the hypothetical Math-152 prerequisite are included in Appendix A. Does transfer-math readiness affect success in CS? Intermediate Algebra is the so-called ‘gate-keeper’ course for transfer level math. If students were required to successfully complete this course before enrolling in these CS courses, would course success increase? Overall, for the eight years studied, the course success rate2 for students in CS-11 is 50.9% and the course success rate for CS-12J is 51.2%. When calculated separately for the group that met the hypothetical prerequisite as opposed to those who did not, students who have completed Intermediate Algebra are more successful. However, an analysis of the variance reveals that the observed increase is statistically significant only for the CS-11 students. The higher levels of course success in CS-12J are possibly due to chance. 3 2 Ratio of students receiving a grade of A, B, C, or Pass to the number enrolled on the census roster. 3 For CS11 p value = .000; for CS-12J p value =.168. Page 4 MATH LEVELS OF COMPUTER SCIENCE Page 5 of 8 Course success rate by Intermediate Algebra completion Hypothetical Algebra prerequisite met? Prereq. met Prereq. not met Prereq. status unknown CS-11 Percent 55.3% 29.5% 35.5% CS12J Total(N) 749 78 107 Percent 56.9% 44.8% 35.0% Total(N) 260 29 80 Though the potential increase for course success among CS-11 students is statistically significant, the explanatory power is weak, explaining only around 3% of the variance4. Other factors, perhaps, have a stronger predictive power. Other factors which influence success in CS Student success can be attributed to a mix of factors. To expand the number of factors considered, other variables such as ethnicity, in the form of under-represented5 or not, and students’ cumulative Cabrillo grade point average (GPA) were added to the statistical model. Both prove to be statistically significant in relation to success in these CS courses, and will account for a slice of the variance.6 Comparatively, a student’s GPA is the strong predictor of success in these CS courses. It is a stronger indicator than completion of Intermediate Algebra. In this model, cumulative GPA accounts for perhaps 17% of the variance in success in CS-11 and around 12% of the variance in CS-12J. This might suggest that good study habits are a better predictor than the level of math preparation. Following this, the next largest accounting for variance in this model is having met the Intermediate Algebra prerequisite. Finally, while underrepresented status has a statistically significant inverse association with success, it is a rather weaker indicator, accounting for just slightly over 1% of the variance in the model.7 4 The R square value is .032 for CS-11. Underrepresented includes Native Americans, African Americans, Filipinos, Hispanic/Latino and Pacific Islanders, with Latino by far the most numerous of the under-represented groups. 6 The p values for GPA are .000 for both CS-11 and CS-12J; the p values for underrepresented status are .000 for CS-11 and .035 for CS-12J. 7 The R square value for GPA is .172 for CS-11 and .119 for CS-12J; the R square value for underrepresented status is .015 for CS-11 and .012 for CS-12J. 5 Page 5 MATH LEVELS OF COMPUTER SCIENCE Page 6 of 8 Taking a broader a look at math levels Thus far, this report has addressed the question of establishing a hypothetical prerequisite, which frames the question in binary terms (yes or no) around a specific level in the math sequence. The weaker than expected explanatory power of meeting or not meeting this hypothetical prerequisite may be related to the fact that the question is very specific. Students at all levels of math are enrolling in CS courses. If the methodology were altered to consider the array of math levels of individual students, would math preparedness emerge as a stronger predictor of success in Computer Science? When a scalar variable for math level8 used in place of the binary met/not met the hypothetical Intermediate Algebra prerequisite, math preparation now emerges as a more powerful predictor of success for students in CS-11 (C++), explaining about 10% of the variance in success. However, this is less so for students in CS-12J (Java), where the explanatory power is about 3.6% of total variance.9 The chart in Appendix B shows success CS-11 and CS-12 according to math level. It is interesting to note that the math level associated with the tipping point – the point where the success in CS-11 and CS12J start to reliably exceed 50%, is one level above transfer-level for CS-11, and transfer-level for CS-12J. Throughout this report, the observed association between math preparedness and success is stronger for C-11 (C++ programming) than for CS-12J (Java programming.) One is tempted to develop hypotheses as to what accounts for this difference. Is the C++ programming language more mathematical? Or are other factors are work? These are questions for future research. Summary Instituting a prerequisite of Intermediate Algebra for CS-11 and CS12J would reduce enrollment by up to 20% in the short term, but would not likely have a disproportionate impact on participation rate of under-represented students or of females. Even though Intermediate Algebra completion is correlated with success in CS-11 (though not 8 Math levels consist of 3 levels below transfer level, transfer or college entry-level, and five other levels which represent the math sequence as determined by the prerequisite hierarchy. For reference, Math-152 (Intermediate Algebra) is one level below transfer. 9 With scalar math levels, the R square values are .100 for CS-11 and .036 for CS-12J. Page 6 MATH LEVELS OF COMPUTER SCIENCE Page 7 of 8 CS-12J), multivariate statistical analysis reveals that Algebra completion has a relatively small predictive value (not significant for CS-12J), and that general academic preparedness is a stronger predictor of success in these courses. When the statistical model is reconfigured to use across-the-board math levels, then a rather strong correlation between math preparation and success emerges for CS-11, less so for CS-12J. Page 7 MATH LEVELS OF COMPUTER SCIENCE Page 8 of 8 Appendix A CS‐11 student ethnicity by math level Hypothetical prerequisite Under‐represented not Met 22.4% 70.1% Not‐met 26.9% 65.4% p= 0.357 CS‐12J student ethnicity by math level Hypothetical prerequisite Female Male Met 20.4% 70.8% Not‐met 31.0% 51.7% p= 0.098 CS‐11 student gender by math level Hypothetical prerequisite Under‐represented not Met 18.3% 81.7% Not‐met 12.8% 87.2% p= 0.230 CS‐12J student gender by math level Hypothetical prerequisite Female Male Met 18.5% 80.8% Not‐met 13.8% 86.2% p= 0.525 Appendix B Success in CS‐11 and CS12‐J by math level Student's math 3 l evels 2 l evel s 1 l evel s Tra ns fer 1 l evel s 2 l evel s 3 l evel s 4 l evel s 5 l evel s bel ow bel ow bel ow Level a bove a bove a bove a bove a bove level Success in CS11 15.4% 31.7% 41.5% 44.6% 66.3% 75.3% 69.1% 82.2% 89.5% Tota l hea dcount (N) Success in CS12J Tota l hea dcount (N) 13 63 188 251 98 77 81 45 19 16.7% 50.0% 45.9% 54.6% 64.5% 73.9% 72.7% 66.7% 100.0% 6 22 74 108 Page 8 31 23 11 12 1