Presentation is loading. Please wait.

Presentation is loading. Please wait.

1 DEVELOPING A MODEL TO EXPLAIN IPEDS GRADUATION RATES AT MINNESOTA PUBLIC TWO- YEAR COLLEGES AND FOUR-YEAR UNIVERSITIES USING DATA MINING For more information.

Similar presentations


Presentation on theme: "1 DEVELOPING A MODEL TO EXPLAIN IPEDS GRADUATION RATES AT MINNESOTA PUBLIC TWO- YEAR COLLEGES AND FOUR-YEAR UNIVERSITIES USING DATA MINING For more information."— Presentation transcript:

1 1 DEVELOPING A MODEL TO EXPLAIN IPEDS GRADUATION RATES AT MINNESOTA PUBLIC TWO- YEAR COLLEGES AND FOUR-YEAR UNIVERSITIES USING DATA MINING For more information contact: Brenda Bailey Ed.D. Associate Director for Research Minnesota State Colleges and Universities Brenda.bailey@so.mnscu.edu

2 2 Minnesota State Colleges and Universities Campus Locations

3 3 Background of the Problem All postsecondary institutions are required to submit the IPEDS Graduation Rate Survey and disclose graduation rates for Student Right-to- Know Reporting graduation rates without reporting supplementary information should be questioned (Astin, 1996) Little is known about using IPEDS data to produce supplementary information about graduation rates at both 2-year and 4-year institutions

4 4 Research Questions 1.What is the relationship between IPEDS graduation rates and institutional characteristics? 2.Given these relationships, what are the predicted graduation rates? 3.How do predicted graduation rates compare to actual graduation rates at Minnesota State system institutions?

5 5 Significance Done at institution level Predicted graduation rates can provide context Little prior research of 2-year college IPEDS data No current research uses data mining on both 2-year and 4-year graduation rates Identified new predictor variables

6 6 Data mining is the process of discovering hidden messages, patterns and knowledge within large amounts of data and making predictions for outcomes or behaviors (Luan, p. 17).

7 7 TRADITIONAL STATISTICAL APPROACH: Deductive Hypothesis Observation Confirmation DATA MINING APPROACH: Inductive Observation Pattern Tentative Hypothesis Theory (Trochim, 2002)

8 8 Fall Collection Winter Collection Spring Collection Institutional Characteristics Survey Completions Survey Employees by Assigned Position Survey Faculty Salaries Survey Fall Staff Survey Enrollment Survey Finance Survey Student Financial Aid Survey IPEDS Peer Analysis System Graduation Rates Survey Data Source: IPEDS Data Collection System

9 9 IPEDS Peer Analysis System Step 1 Download IPEDS Data Microsoft Excel Files Step 2 Build Data Mining Files Microsoft Access and SPSS Software Step 3 Data Mining C&RT Clementine Software Weighted Predicted IPEDS Graduation Rates Microsoft Access Flow Chart of Data Analysis

10 10 Algorithm Classification and Regression Tree (C&RT) Tree-based classification and prediction method with binary splits Examines input fields and splits records into peer groups with similar output field values Graduation rate was set as the output variable All other IPEDS variables were set as input fields Variables can be nominal or ordinal (categorical) or interval (scale) Predicted graduation rate is the average graduation rate for each peer group The researcher also calculated a weighted predicted graduation rate for the institutions in each peer group.

11 11 ModelCount Pearson Correlation r 1Private for-profit four-year2110.885 2Public four-year5860.877 3Public two-year and less1,4210.854 4Private not-for-profit less than 2-year1140.846 5Private not-for-profit two-year only2210.817 6Private not-for-profit four-year1,2730.754 7Private for-profit two-year only7220.751 8Private for-profit less than two-year1,2230.672 Total5,771 Strong Relationship Between Actual and Predicted Graduation Rate

12 12 SurveyCount Enrollment22 Institutional Characteristics19 Student Financial Aid2 Graduation Rate2 Salaries2 NPEC-Salaries1 Staff1 NPEC-Finance1 Completions1 Total51 Source of Predictor Variables

13 13 Private For-Profit Four-year Model Predictors First Split1. Percent of enrollment that is men 2. Carnegie Classification Code 3. Enrollment age 20-21 total 4. First-time, degree-seeking enrolled PT women 5. Full-year unduplicated graduate HC Non-Resident Alien 6. Full-year unduplicated undergrad headcount Hispanic 7. Percent of enrollment that is first-time 8. Percent of enrollment that is first-time men 9. Service/maintenance staff men NEW 10. State of institution 11. State of residence when student was first admitted 12. Total Awards: Computer and Information Sciences NEW 13. Total completers within 150% of normal time 14. Tuition plan restricted NEW

14 14 First Split1. Percent of full-time enrollment that is White 2. Average faculty salary male NEW 3. Average faculty salary professor male NEW 4. Enrollment American Indian 5. First-time, degree-seeking enrolled part-time men 6. Full-time enrollment women 7. Full-time retention rate 8. Full-year unduplicated headcount women 9. % of scholarship expenditures from Pell grants 10. % of first-time degree-seeking students submitting SAT 11. % receiving institutional grant aid 12. SAT 1 Math 75th percentile score 13. State of institution 14. Total completers within 150% of normal time 15. Total dormitory capacity Public Four-year Model Predictors

15 15 First Split1. Highest Degree offered 2. Adjusted cohort 3. Enrollment age 18-19 women 4. Enrollment age 20-21 women 5. Enrollment age 22-24 men 6. Total completers within 150% of normal time Public Two-year and Less Model Predictors

16 16 Private Not-for-Profit Less than 2-year Model Predictors First Split1. Regional accrediting agency NEW 2. Books and supplies in largest program NEW 3. CIP Code of largest program NEW 4. Degree of urbanization 5. Full year undergraduate White enrollment 6. Full-time Black enrollment 7. Offers programs not leading to a formal award NEW 8. State of institution 9. Total completers within 150% of normal time

17 17 Private Not-for-Profit Two-year Only Model Predictors First Split1. 12-month instructional activity credit hours: undergrad 2. Average amount of institutional grant aid received 3. Calendar system NEW 4. Current year GRS cohort as a % of entering class NEW 5. Full year undergraduate White enrollment 6. None of the special learning opportunities are offered NEW 7. Off campus not with family other expenses NEW 8. Off campus with family other expenses NEW 9. Percent of full-time enrollment that is men 10. Percent of undergraduate enrollment that is Black 11. State of institution 12. Total completers within 150% of normal time

18 18 Private Not-for-Profit Four-year Model Predictors First Split1. Carnegie Classification Code 2. Adjusted cohort 3. Average faculty salary total NEW 4. Full-time retention rate 5. Name of Regional accrediting agency NEW 6. SAT I Math 25 th percentile score 7. Total completed within 150% of time

19 19 Private For-Profit Two-year Only Model Predictors First Split1. Total completers within 150% of normal time 2.Adjusted cohort 3.Full-year undergraduate total enrollment 4.State abbreviation code of institution Private For-Profit Less than Two-year Model Predictors First Split1. Total completed within 150% of normal time 2. Adjusted cohort

20 20 Minnesota State System Four-year Predictor Variables Differ by Group GroupPredictor Variables 8% White Completers Full-time Women Room Capacity % Pell Expenditures 12% White Completers Full-time Women Room Capacity % Submitting SAT 5% White Completers % with Grant Aid Unduplicated Headcount Women American Indian Enrollment 3% White Completers % with Institutional Grant Aid State 11% White Male Faculty Professor Salary Male Faculty Salary Retention State

21 21 Minnesota State SystemTwo-year Predictor Variables Differ by Group GroupPredictor Variables 8Highest Degree Women 18-19 Completers Women 20-21 3Highest Degree Women 18-19 Completers Women 20-21 Completers 6Highest Degree Women 18-19 Completers Women 20-21 Completers 1Highest Degree Women 18-19 Completers Cohort 7Highest Degree Women 18-19 Completers Cohort 11Highest Degree Women 18-19 Completers Cohort

22 22 Models Compared to Current Methodology ModelCount Pearson Correlation rRelationship 2Public four-year5860.877Strong Current Minnesota State system four-year method5860.603Medium 3Public two-year and less1,4210.854Strong Current Minnesota State system two-year method1,4210.675Strong

23 23 Some New Predictors Average male faculty salary Number of awards in Computer Science Number of service/maintenance men Regional accrediting agency No special learning opportunities offered CIP code of largest program Cost of books and supplies in largest program Calendar system Other expenses off campus GRS cohort as a percent of entering class

24 24 So What at Minnesota State System? Could provide national context for Student-Right-to- Know Disclosure forms Could provide national context for graduation rate reports and accountability measures Identifies peers groups for Minnesota State system colleges and universities Shows different predictors for different sectors and peer groups within the system Data mining techniques could be used for other system research projects


Download ppt "1 DEVELOPING A MODEL TO EXPLAIN IPEDS GRADUATION RATES AT MINNESOTA PUBLIC TWO- YEAR COLLEGES AND FOUR-YEAR UNIVERSITIES USING DATA MINING For more information."

Similar presentations


Ads by Google