Upload
lester-hunt
View
215
Download
2
Embed Size (px)
Citation preview
MENOAn admission test experiment at the Vrije Universiteit Brussel
Raoul Van Esbroeck
23/11/2006 2
Vrije Universiteit Brussel
Pag.
Why an admission test in Flanders?
• Open admission• High failure rate• Costs for society• Costs for the individual student• Need for more efficiency• Admission test a possible option
23/11/2006 3
Vrije Universiteit Brussel
Pag.
Why the MENO-test?
• UCLES (University of Cambridge Local Examination Syndicate)
• Skills crucial to success in HE• Limited to 50 multiple choice items
• Critical Thinking (CT) with 3 subscales• Problem Solving (PS) with 3 subscales
• Good reliability and validity• Limited bias (language, age, ethnic origin)• Gender and SES bias?
23/11/2006 4
Vrije Universiteit Brussel
Pag.
Admission test experiment
• OBPWO project (Promoter: Prof. M. De Metsenaere)
• Questionnaire• Social and cultural variables• MENO- test
• Administered in class during first week • Cohort of entering Freshmen in 95-96• Respondents: 85% of the target group (N =
1000)
23/11/2006 5
Vrije Universiteit Brussel
Pag.
Does MENO allow correct classification?
• Errors in classification• False positives (FP)• False negatives (FN)
• Total number of errors% refused %FN %FP10 6 8725 18 6350 39 39
• Conclusion: too many errors
23/11/2006 6
Vrije Universiteit Brussel
Pag.
Has the test an added value?
• Can the test predict academic success better than SE results?
• A model with 4 variables:• SE option, class rank, graduating on time• MENO
• Explains 18.7% of variance• MENO represents only 1.75% in this model
• Conclusion: only limited added value
23/11/2006 7
Vrije Universiteit Brussel
Pag.
Bias?
• There is a gender bias • Girls would be more refused than boys• FN error is higher with girls• FP error is higher with boys
• There is a SES bias• Lower SES will be more refused than higher SES
groups• FN error is higher with lower SES groups• FP error is higher with higher SES groups
23/11/2006 8
Vrije Universiteit Brussel
Pag.
Increase of success rate?
• There is limited increase
• If 25% of the lowest MENO-scores are eliminated success rate increases from
49% to 58%
• What about errors?
23/11/2006 9
Vrije Universiteit Brussel
Pag.
Conclusion on MENO
• MENO is not an acceptable admission test• Too many errors in classification • Limited added value• There is gender and SES bias• The gains are limited
• Using this test would be unethical
23/11/2006 10
Vrije Universiteit Brussel
Pag.
Are other tests better?
• There are indications that the MENO in the UK is also biased
• The Chrysostomos test (P. Janssen, KULeuven) also
faces classification errors• A “Pass” (50%) score on the test leads to 37.6% of classification errors• A 60% score on the test leads to 30.9% of classification errors (with
about 5% of FN)
• Similar problems exist in other countries where admission tests are used
23/11/2006 11
Vrije Universiteit Brussel
Pag.
General conclusion
• There is no need for admission tests because • There is limited added value• Too many errors are made • There is bias• The gains in success rate are limited• The costs for the tests are extremely high
• There is no real societal benefit if FN are to be avoided
• There will be too many wrong decisions if you want to have real gains
23/11/2006 12
Vrije Universiteit Brussel
Pag.
Personal reflection
• The use of general admission tests for selection is unacceptable and unethical
• Such tests may be excellent for guidance purposes