This study evaluated the relationship between student use of a Learning Management System (LMS), student characteristics, and academic achievement in a large undergraduate hybrid course. The results showed:
1. LMS use, particularly assessment activities, had a stronger correlation with course grades than student characteristics like high school GPA.
2. A regression model combining LMS usage data and student characteristics moderately predicted student success, suggesting refinement could strengthen results.
3. LMS data required extensive filtering to be useful for analysis, and student variables needed screening for missing data.
The conclusions were that LMS behavioral data may better predict outcomes than demographics alone, and combining LMS and student data in a complete model relates to academic achievement
1. Logging On To Improve Achievement
Evaluating the relationship between use of the Learning
Management System, student characteristics, and academic
achievement in a hybrid large enrollment undergraduate course
Research Highlights: Presentation to SoLAR Storm
November 15, 2012
John C Whitmer (jwhitmer@calstate.edu)
Committee Chair: Dr. Paul Porter, Sonoma State University
Slides:
http://slidesha.re/sFKjcm
2. Introduction
• Educational Doctorate
Degree (EdD) candidate
(University of California
Davis & Sonoma State
University)
• Advanced to candidacy,
defending ~ January 14
• Associate Director,
California State University
LMSS Project,
Chancellor’s Office
3. Presentation Outline
1. Study Case & Context
2. Results for Instructional Practices
3. Results for LMS Data Analysis
4. Conclusions
5. Problem: Student Graduation
• Less than 50% of college/university students graduate
within 6 years
• California State University: 52.4%
(first-time freshman, 2000 cohort)
(CSU Analytic Studies, 2011)
• Students from under-represented minority racial/ethnic
groups graduate at much lower rates
• California State University: 38.3%
(African American students, first-time freshman, 2000 cohort)
(CSU Analytic Studies, 2011)
• Contributing factor: mega-enrollment intro courses
• Infrequent interaction, prevent faculty/student relationships
6. Case: Introduction to Religious Studies
• Redesigned to hybrid 54 F’s
delivery through Academy
eLearning
• Highest LMS usage entire
campus Fall 2010
(>250k hits)
• 373 students (54% increase)
• Bimodal results
• 10% increased SLO mastery
• 7-11% increase in DWF
7. Research Questions
1) Is there a relationship between student LMS usage and academic
performance? Does this relationship vary by the pedagogical
purpose underlying LMS usage? (correlation)
2) Is there a relationship between student background characteristics
or current enrollment information and academic performance?
(correlation)
3) Does analyzing combined student characteristics and current
enrollment information increase the predictive relationship between
combined LMS usage data and student success?
(multivariate regression)
4) Does a student’s economic status and student of color status
vary the predictive relationship between combined LMS usage,
combined background characteristics and current enrollment
information? (multivariate regression, restricted model)
8. Independent Variables: Student Characteristics
1. Gender
2. Under-Represented Minority
3. Pell-Eligible
4. High School GPA
5. First in Family to Attend College
6. Major-College
7. Enrollment Status
Under-Represented Minority and Pell-Eligible
8. (interaction)
Under-Represented Minority and Gender
9. (interaction)
10. Research Methods (Cliff’s notes version)
1. Extract data, validate with appropriate “owner”
2. Transform variables
• measures of interest (e.g. “URM”, not race/ethnicity)
• analysis methods (categorical into numeric)
3. Examine data for
• outliers, missing data, data distributions, etc.
• colinearity between variables (e.g. independence)
4. Join data into single data file, collapse to one record/student
5. Run analysis
21. Conclusions
1. At the course level, LMS use better predictor of
academic achievement than any student characteristic
variable. Behavioral data appears to supercede
demographic information (what do, not who are).
2. Moderate strength magnitude of complete model
demonstrates relevance of data, but suggests that
refinement of methods could produce stronger results.
3. LMS data requires extensive filtering to be useful;
student variables need pre-screening for missing data.
22. Ideas & Feedback
Potential for improved LMS analysis methods:
• social learning
• activity patterns
• discourse content analysis
• time series analysis
Group students by broader identity, with unique variables:
• Continuing student (Current college GPA, URM, etc.
• First-time freshman (HS GPA, SAT/Act, etc)