Case #8 PREDICTICTING GRADUATION RATES FOR COLLEGES AND UNIVERSITIES A group work presented to the faculty of the Master

r of Business Administration Far Eastern University

In partial fulfilment of the requirements In Managerial Statistics For the degree of Master of Business Administration

Submitted to: Prof. Edgardo Montesclaros

Submitted by: Group 2 Lomibao, Myra Joy M. Sebastian, Regie Abel R. Rabang, Gracela C.

APRIL 10, 2013

CASEPROBLEMPREDICTICTING GRADUATION RATES FOR COLLEGES AND UNIVERSITIES The percentage of students who enroll at a college or university and actually graduate are an important statistic for university administrators. Some of the factors related to the graduation rate include the percentage of classes with fewer than 20 students, the percentage of classes with more than 50 students, the student-faculty ratio, the percentage of students who apply to the university and are admitted, the percentage of first-year students in the top 10% of their high school class, and the academic reputation of the university. To study the effect of these factors on the graduation rate, data for 48 national universities was collected ( Americas Best Colleges, Year 2000 Edition). Descriptions for the data follow: State Region Graduation Rate % of Class Under 20 % of Classes of 50 or More Student-Faculty Ratio Acceptance rate 1st-Year Students in Top 10% of HS Class Academic reputation Score The state in which the university is located The region of the country in which the university is located (North, South, Midwest, West) The percentage of students who enroll at the university and graduate The percentage of classes with fewer than 20 students The percentage of classes with more than 50 students The ration of the number of students enrolled divided by the total number of faculty The percentage of students who apply and are accepted The percentage of students admitted who were in the top 10% of their high school class A measure of the schools reputation determined by surveying administrators and other universities: measured on a scale form 1 (marginal) to 5 (distinguished)

CASEPROBLEMMANAGERIAL REPORT Use the methods presented in this and previous chapters to analyze this data set. Present a summary of your analysis, including key statistical results, conclusions, and recommendations in a managerial report. Include any appropriate technical material (computer output, residual plots, etc.)

CASEPROBLEMDATA SET
% of Classes Under 20 65 73 65 68 77 69 68 64 65 69 . . . 32 32 53 56 45 39 37 42 31 % of Classes of 50 or More 7 9 16 11 9 13 7 15 8 8 . . . 22 28 17 8 20 18 17 22 24 Student/ Faculty Ratio 3 8 6 5 7 7 8 9 7 7 . . . 19 19 13 12 20 21 12 20 23 1st-Year Students in Top 10% of HS Class 100 90 95 93 95 87 89 81 90 87 . . . 48 95 61 47 95 44 37 90 56 Academic Reputation Score 4.7 4.9 4.9 4.8 4.8 4.9 4.6 4.7 4.4 4.6 . . . 3.9 3.8 3.6 3.5 3.5 4 3.9 3.6 3.6

Observation 1 2 3 4 5 6 7 8 9 10 . . . 40 41 42 43 44 45 46 47 48

School California Institute for Technology Harvard University Massachusetts Inst. Of Technology Princeton University Yale University Stanford University Duke University John Hopkins University University of Pennsylvania Columbia University . . . Pennsylvania State Univ. University of California-Davis Univ. of Southern California Tulane University Univ. of California-Santa Barbara University of Texas-Austin University of Washington University of California-Irvine University of Florida

State CA MA MA NJ CT CA NC MD PA NY . . . PA CA CA LA CA TX WA CA FL

Region West North North North North West South South North North . . . North West West South West West West West South

Graduation Rate 85 97 92 95 94 92 92 89 90 90 . . . 80 74 70 72 70 66 70 74 67

Acceptance Rate 18 12 22 13 18 13 28 41 29 14 . . . 47 66 45 79 61 71 66 62 61

MODELBUILDINGCORRELATION ANALYSIS 1st-Year % of Students Student/Faculty Acceptance Classes of in Top Ratio Rate 50 or More 10% of HS Class

Graduation Rate Graduation Rate % of Classes Under 20 % of Classes of 50 or More Student/Faculty Ratio Acceptance Rate 1st-Year Students in Top 10% of HS Class Academic Reputation Score 1 0.583 -0.457 -0.605 -0.799 0.588 0.631

% of Classes Under 20

Academic Reputation Score

1 -0.690 -0.786 -0.571 0.420 0.461

1 0.748 0.301 0.037 -0.132

1 0.482 -0.250 -0.451

1 -0.658 -0.622

1 0.507 1

In this correlation analysis, we only consider the first column, which shows the relation between the graduation rate and the other explanatory variables. We can see that the value between graduation rate and graduation rate is equal to 1, indicating a strong relationship, since they are the same variable. The negative values show a negative correlation, meaning that if the graduation rate increases, the other variable will decrease (and vice versa). The positive values show a positive correlation, meaning that if one increases, the other will also increase (and vice versa). Therefore, the % of Classes of 50 or More, Student/Faculty Ratio, and Acceptance Rate have a negative correlation with Graduation Rate. The weakest relationship is between graduation rate and the % of Classes of 50 or more, with a correlation coefficient of 0.457. The strongest relationship is between graduation rate and acceptance rate, with a correlation coefficient of 0.799.

MODELBUILDINGVARIANCE INFLATION FACTOR Adjusted R2 0.7080 0.6869 0.7028 0.5735 0.5267 0.4353 Standard Error 7.1293 3.5161 2.6443 12.3502 11.8692 0.3588 Variance Inflation Factor (VIF) 3.8325 3.5743 3.7658 2.6240 2.3643 1.9816

Variable % of Classes Under 20 and all other X % of Classes of 50 or More and all other X Student/Faculty Ratio and all other X Acceptance Rate and all other X 1st-Year Students in Top 10% of HS Class Academic Reputation Score and all other X

Multiple R 0.8597 0.8487 0.8570 0.7867 0.7596 0.7038

R2 0.7391 0.7202 0.7345 0.6189 0.5770 0.4953

Since there are no variance inflation factors (VIFs) greater than 5, there is no evident multicollinearity between all independent variables. All independent variables are relatively uncorrelated. This allows for further testing and determination of the best model among class of models.

MODELBUILDINGSTEPWISE ANALYSIS Acceptance Rate entered. df Regression Residual Total 1 46 47 SS MS F 2228.019659 2228.019659 81.73630187 1253.897008 27.2586306 3481.916667 Standard Error t Stat 1.866173454 52.76927939 0.040269375 -9.04081312 Significance F 9.08645E-12

Intercept Acceptance Rate

Coefficients 98.476628 -0.3640678

P-value Lower 95% 8.35305E-43 94.72021612 9.08645E-12 -0.44512594

Upper 95% 102.2330407 -0.283009847

Student/Faculty Ratio entered. df Regression Residual Total 2 45 47 SS MS F 2445.832658 1222.916329 53.11464548 1036.084009 23.02408908 3481.916667 Standard Error 2.0042944 0.042249 0.1647164 Significance F 1.43045E-12

Intercept Acceptance Rate Student/Faculty Ratio

Coefficients 101.66652 -0.3013868 -0.5066263

t Stat 50.724347 -7.1335086

P-value 2.34229E-41 6.46186E-09

Lower 95% 97.629671 -0.386481 -0.8383823

Upper 95% 105.703383 -0.2162921 -0.1748702

-3.0757474 0.003565133

No other variables could be entered into the model. Stepwise ends.

90 80 70 60 50

Relationship of Graduation Rate vs. Student/Faculty Ratio and Acceptance Rate

40
30 20 10 0 0 20 40 60 80 100 120

This scatter plot shows that there is a negative linear relationship between Graduation Rate and Acceptance Rate, with a steep downward slope. Since the slope is steep, this means that the rate of change of graduation rate will be less, for every change in acceptance rate. There is also a negative linear relationship between Graduation Rate and Student-Faculty Ratio, with a flatter downward slope. This indicates a more rapid rate of change of graduation rate for every change in the student-faculty ratio.

STEP1FORMULA&SLOPE

y = a + b1x1 + b2x2 y = 101.67 - 0.30x1 - 0.51x2

The prepared formula proves the conclusion of the earlier scatter plot. For each unit increase in acceptance rate, expect graduation rate to decrease by 0.30, assuming that the student-faculty ratio is held constant. For each unit increase in student-faculty ratio, expect graduation rate to decrease by 0.51, assuming that the acceptance rate is held constant. STEP2REGRESSION
Regression Statistics Multiple R R Square Adjusted R Square Standard Error Observations 0.8381161 0.702438597 0.689213646 4.798342326 48

With an R2 equal to 0.7024, this means that 70.24% of the variability of graduation rates is explained by the acceptance rate and the student-faculty ratio. The adjusted R2 is equal to 68.92, which is significantly close to the computed R 2, indicating that we can expect minimal shrinkage when generalizing to the population.

STEP3PREDICTIVENESS

Intercept Acceptance Rate Student/Faculty Ratio

Coefficients Standard Error 101.666527 2.0042944 -0.30138689 0.0422494 -0.50662633 0.16471649

t Stat 50.724347 -7.1335086 -3.0757474

P-value 2.34229E-41 6.46186E-09 0.00356513

Lower 95% 97.629671 -0.38648167 -0.8383823

Upper 95% 105.703383 -0.2162921 -0.1748702

Step 1 Ho: Ha: Rejection Rule p-value Decision Conclusion

Is Intercept Significant? =0 0 0.05 Reject Ho & Accept Ha if p< 0.00 Reject Ho & Accept Ha The intercept is significant.

Is Slope of Acceptance Rate Significant? =0 0 0.05 Reject Ho & Accept Ha if p< 0.00 Reject Ho & Accept Ha The slope of Acceptance Rate is significant.

Is Slope of StudentFaculty Ratio Significant? =0 0 0.05 Reject Ho & Accept Ha if p< 0.00 Reject Ho & Accept Ha The slope of StudentFaculty Ratio is significant.

Is Model Predictive? Model is not predictive Model is predictive 0.05 Reject Ho & Accept Ha if p< 0.00 Reject Ho & Accept Ha The model is PREDICTIVE.

2 3

4 5

MAJOR CONCLUSION: Since the model is predictive, then the acceptance rate and the student-faculty ratio significantly both affect the graduation rate.

STEP4ADEQUACY Residual Analysis for Linearity and Homoscedasticity

Acceptance Rate Residual Plot

Student/Faculty Ratio Residual Plot

10

15

20

25

Student/Faculty Ratio

Both plots appear linear and homoscedastic. This means that the probability distributions of error for acceptance rate and studentfaculty ratio are normal. Also, there is constant variance for the residual plots. Durbin Watson Analysis Durbin-Watson Statistic DU (n=50, k=2) DL (n=50, k=2) 2.269 1.628 1.462

Since the Durbin-Watson statistic is above the upper value, we can conclude that there is no statistical evidence that the error terms are autocorrelated. Furthermore, since the Durbin-Watson statistic is greater than 2, we can conclude independence of errors. The model is ADEQUATE.

MODELBUILDINGSUMMARY OF RESULTS&FINAL CONCLUSIONS After performing the stepwise regression analysis, we determined that the best explanatory variables to use for the model are Acceptance Rate and Student-Faculty Ratio. These two independent variables are chosen to explain the Graduation Rate (dependent variable) Moreover, in the regression analysis, we conclude that the model is predictive and adequate. From the scatter plots of the relationship between Graduation Rate versus Acceptance Rate and Student-Faculty Ratio, we determined that there is a linear relationship between the dependent and independent variables. Since both of the slopes are downward, we can conclude that the Acceptance Rate and Student-Faculty Ratio have a negative relationship. Therefore, if either the Acceptance Rate or Student-Faculty Ratio increases, then the Graduation Rate will decrease.

ACCEPTANCE RATE

STUDENT-FACULTY RATIO

GRADUATION RATE

MODELBUILDINGRECOMMENDATIONS Since the model is adequate, the administrators of major universities should manage and control the number of enrollees. A large number of students and a high student-faculty ratio may significantly affect the concentration of the student. Additionally, if there is a considerably large class being taught by one professor, there is less focus on the students individually. University Administrators could possibly create a more intricate application procedure to minimize the number of enrollees.