PAR Framework Review and Summary
Ellen Wagner, chief strategy officer, PAR Framework
Beth Davis, managing directorPAR Framework
Data Are Changing Everything
Data and Evidence basedDecision making in Higher Ed
Analytics have ramped up everyone’s expectations ofpersonalization, accountability and transparency.Academic enterprises cannot live outside theinstitutional focus on tangible, measurable resultsdriving IT, finance, recruitment and other missioncritical concerns.
Costs and Completion Rates
Source: New York Times; NCES
0
10
20
30
40
50
60
70
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2 yr colleges4 yr colleges
Graduation rates at 150% of time
Cohort year
Performance Based Funding
http://www.ncsl.org/issues research/educ/performance funding.aspx
Institutional Accountability
http://www.whitehouse.gov/issues/education/higher education/college score card
While Big Data raise expectations,student data drive big decisions in .edu
PAR Framework video
• https://www.dropbox.com/s/jksxbc6uac4n1rg/PAR_pre delivery.mp4
PAR Outputs
IdentifyShow how institutionscompare to their peers instudent outcomes, byscaling amultiinstitutional databasefor benchmarking andresearch purposes.
Target
Identify which studentsneed assistance, by usingin depth, institutionalspecific predictive models.
Models are unique to theneeds and priorities of ourmember institutions basedon their specific data.
Determine best ways toaddress weaknessesidentified in benchmarksand models by scalingand leveraging amember, data andliterature validatedframework for examininginterventions within andacross institutions(SSMx)
Treat
PRELIMINARY UND FINDINGS FROM PAR, SPRING, 2014
Beth DavisManaging DirectorPAR Framework
Rapid Results
• Data Delivery– First data meeting 1/24– Preliminary data provided by 2/18– Final data expected 3/24
• Early discovery– UND course catalog– Basic student information file– Did not include financial aid, credentials
With limited data set and in less than 1 week
Identified• The key gatekeeper courses• With over 70% accuracy the likelihoodstudents will succeed in college level courses *
• 10 most predictive factors known aboutstudent at entry.
* Accuracy will increase when data is complete, reaches 80 90+% for many scenarios
Predictions
• Everyone who started fall 2013.• A prediction of them taking one of thecourses.
• Representative of early success in a collegelevel course.
Course name and course success Key factors include:HS GPAHS credits by examTransfer CreditsRaceGender
HS GPA >3.8
HS GPA < 2.6
Top Risk Factors
• HS GPA less than 2.65• GED• Race (B, AI, H)• Age at start ><19• Gender• Transfer status
Demographics that impact student success
Risk Factors • Students who had a GPA between 0 and 2.66failed the gateway courses 57% of the time.
Risk ratios • The factors where variation impacts studentsuccess and by how much
• Students with a HS GPA not specified are 4X lesslikely to succeed in college level courses
WatchlistPARanonymized ID 1st, 2nd and 3rd most important
factors contributing to risk
Risk they will notsucceed in collegelevel course
DISCUSSION
1229Thomas M. DiLorenzo, Ph.D. • Exceptional UND: A Critical Step in Realizing the Vision April 22, 2014 • University of North Dakota
Performance metricsInformed decision-making forStudent Success Systems
1230Thomas M. DiLorenzo, Ph.D. • Exceptional UND: A Critical Step in Realizing the Vision April 22, 2014 • University of North Dakota
OUR APPROACH
• Progress can be made quickly• Graduation rates are everyone s responsibility• We need data-based approaches and• Proactive/prevention systems• Focus part of your time on students through systems and
processes that provide efficient, accurate, and integrated support
• And part of your time on at-risk students, identified through custom systems
1231Thomas M. DiLorenzo, Ph.D. • Exceptional UND: A Critical Step in Realizing the Vision April 22, 2014 • University of North Dakota
EFFICIENT, ACCURATE AND INTEGRATED SYSTEMS
• Online Catalog• Standard Schedule• Communications Strategy• Degree Audit/Planner• One Stop Student Services Center
1232Thomas M. DiLorenzo, Ph.D. • Exceptional UND: A Critical Step in Realizing the Vision April 22, 2014 • University of North Dakota
SYSTEMS/STRUCTURAL CHANGES (EXAMPLES)
• Major selection• 120-90-75-60 Initiative• Four year graduation plans
1233Thomas M. DiLorenzo, Ph.D. • Exceptional UND: A Critical Step in Realizing the Vision April 22, 2014 • University of North Dakota
PROACTIVE INTERVENTIKONWITH AT-RISK STUDENTS
• Faculty Early Alert System• Advisor CRM (Client Relationship
Management)• Advisor Utilization• Departed Student Outreach• Advanced Data Analysis
1234Thomas M. DiLorenzo, Ph.D. • Exceptional UND: A Critical Step in Realizing the Vision April 22, 2014 • University of North Dakota
ASSESS AT-RISK /DEVELOP INTERVENTIONS
• Partnership between Academic Affairs, Student Affairs and Financial Affairs
• Faculty involvement in all phases• Student involvement in all phases• Dean and Chairs involvement in all phases
1235Thomas M. DiLorenzo, Ph.D. • Exceptional UND: A Critical Step in Realizing the Vision April 22, 2014 • University of North Dakota
1236Thomas M. DiLorenzo, Ph.D. • Exceptional UND: A Critical Step in Realizing the Vision April 22, 2014 • University of North Dakota
EXAMPLES
• Undergraduate research• Internships• Service learning• Study abroad• Undergraduate scholarships/fellowships• Honors experiences
The Predictive Analytics Reporting (PAR) Framework
• PAR is a national, non profit multi institutionalcollaborative focused on institutional effectiveness andstudent success.
• PAR is a “big data” analysis effort using predictiveanalytics to identify drivers related to loss and momentumand to inform student loss prevention
• PAR member institutions voluntarily contribute deidentified student records to create a single federateddatabase.
• Descriptive, inferential and predictive analyses have beenused to create benchmarks, institutional predictive modelsand to map student success interventions to predictorbehaviors
Common Definitions Lead to Shared Understanding
Analysis/Modeling Process
• Analysis and model building is aniterative process
• Around 70 80% efforts are spenton data exploration andunderstanding.
Structured, Readily Available Data• Common data
definitions = reusablepredictive models andmeaningfulcomparisons.
• Openly published via acc license @https://public.datacookbook.com/public/institutions/par
PAR Data Inputs Student
Demographics& Descriptive
GenderRace
Prior CreditsPerm Res Zip CodeHS InformationTransfer GPAStudent Type
Student CourseStudent CourseInformationCourse Location
SubjectCourse Number
SectionStart/End DatesInitial/Final GradeDelivery ModeInstructor StatusCourse Credit
StudentAcademicProgress
Curent Major/CIPEarned Credential/CIP
StudentFinancial
InformationFAFSA on File – Date
Pell Received/Awarded –Date
Course CatalogSubject
Course NumberSubject LongCourse Title
Course DescriptionCredit Range
** Future
Lookup TablesCredential Types OfferedCourse Enrollment Periods
Student TypesInstructor StatusDelivery ModesGrade Codes
Institution Characteristics
Possible Additional **Placement TestsNSC InformationSES Information
Satisfaction SurveysCollege Readiness SurveysIntervention Measures
PAR OutputsDescriptiveBenchmarks
Show how institutionscompare to their peers instudent outcomes, byscaling amultiinstitutional databasefor benchmarking andresearch purposes.
PredictiveModels
Identify which studentsneed assistance, by usingin depth, institutionalspecific predictive models.Models are unique to theneeds and priorities of ourmember institutions basedon their specific data.
Institutions addressareas of weaknessidentified inbenchmarks and modelsby scaling and leveraginga member, data andliterature validatedframework for examininginterventions within andacross institutions(SSMx)
InterventionMatrix
Feedback loops for enabling institutional performance improvements
PerformanceBenchmarksPerformanceBenchmarks
InterventionBenchmarksInterventionBenchmarks
PredictiveModelsPredictiveModelsActionAction
MeasurableResults
MeasurableResults
CommonData
Definitionsand DataWarehouse
Scalable cross institutional improvements enabled byCollaboration via PAR
Benchmarking in risk factor areas
PAR Student Success Matrix (SSMx)
Literature based tool forbenchmarking student services
and interventions 600+ total interventions submittedAbility to compare among all 16 PAR institutions Basis for institutional intervention field tests Publically available, over 1,000 downloads since June 2013
https://par.datacookbook.com/public/institutions/par
Student Success Framework Identified 38 distinct functionalcategories of interventions
• Mapped >600 interventionsto functional categories foreasy benchmarking andcomparisons
Literature and partner validatedpredictors reveal 80 risk factors
• Tying interventions to thepredictive factors enablesinsight into how to bestapply institutional resources
• level of intervention (student, course, section, program,institution, . . .)
• focus (audience)
• delivery channels
• impact (# students affected)
• results measurement (outcomes, if available)
• Return on Investment (ROI)
©PAR Framework 2013
SSMX ONLINE APPLICATION
Summary, Conclusions
Thank You