Predicting Individual Student Attrition and Fashioning Interventions to Enhance Student Persistence

1
Predicting Individual Student Attrition and
Fashioning Interventions to Enhance Student
Persistence and Success

• Dr. Thomas E. Miller
• University of South Florida

2
Introduction

• Sources of concern for persistence and graduation
  rates
• - institutions all component parts
• - government
• - college rating services
• - public
• Common approaches have been broadly implemented
• - generally targeted to sub-populations
• - necessarily inefficient and wasteful as
  persistence
• enhancement tools
• - may still be sound practice for educational
  reasons

3
Introduction cont.

• This project is specific to each student based on
  established weighted predictors
• - allows for timely response (uses
  pre-matriculation data)
• - efficient
• - replicable
• - responsive to individual needs and interests

4
Background

• Pascarella and Terenzini (1980) applied Tintos
  model of social integration.
• - findings valued the interaction between
  students and faculty
• - addressed post-matriculation issues
• Chapman and Pascarella (1983) studied
  social and academic integration.
• findings revealed differences in levels of social
  and academic integration.

5
Background cont.

• Canisius College model predicted attrition for
  specific students.
• - successful, still used
• - freshman to sophomore persistence rate
• - graduation rates
• - variables in logistic regression formula
  included
• high school average
• gender
• academic behaviors in high
  school
• parents together

6
CSXQ

• Normally used to compare how students
  expectations for college align with their actual
  experiences
• For this study CSXQ data are examined to
  determine their worth in predicting student
  persistence.
• Supplemental data such as gender, ethnicity, age,
  academic performance potential will be used along
  with the CSXQ data in the predictive model.

7
(No Transcript)
8
(No Transcript)
9
(No Transcript)
10
(No Transcript)
11
Methodology

• The CSXQ was administered to First Time in
  College (FTIC) freshman prior to matriculation in
  the fall of 2006. Participants were 3,998
  student on Tampa campus
• Slightly fewer than 1,000 completed the survey
  and gave identifying information
• The sample was representative of the larger
  population in every demographic measure.

12
Results

• The PROC LOGISTIC procedure in SAS was run using
  set-wise inclusion of variables.
• Two blocks of independent variables dependent
  variable persist/not persist
• Block One
• Block Two

13
(No Transcript)
14
(No Transcript)
15
(No Transcript)
16
Predicting New Cases

• Focusing on Block Two variables, predictors are
• 1. High School GPA ()
• 2. Being Black vs being white ()
• 3. Expecting to participate in clubs/student
  organizations ()
• 4. Expecting to read many textbooks or assigned
  books in college ()
• 5. Expecting to read many non-assigned books in
  college (-)
• 6. Expecting to work off campus while in
  college (-)

17
Other variables that may prove useful

• Institutional data
• - Gender
• - Honors Program
• - Early enrollment summer programs
• - Residence
• - Number of guests at summer orientation
• - Date of summer orientation program
• - Date of application for admission
• - Permanent residence out of state
• - Major is pre-nursing or pre-education

18
Other variables cont.

• CSXQ data
• - plan to be employed on campus
• - intended effort scale related to course
  learning
• - intended effort scale related to scientific and
  quantitative experiences.

19
Interventions

• Model will identify approximately 500-700 FTIC
  students at risk of attrition in their first
  year, of the total 4,200 enrolled.
• Levels of intervention
• Employing those already interacting with student
  sub-sets
• - athletics
• - undeclared majors
• - residence halls
• - summer programs
• - Honors College
• - Others First-Year Student Connections

20
Referral points (post-intervention)

• Career services
• Academic advising
• Financial aid
• Residence halls

21
Conclusion Next Steps

• Model refinement
• Increasing and refining interventions
• Predicting sophomore persistence
• Transfer students