Test separation value

from Wikipedia, the free encyclopedia

A test separation value , threshold value or limit value ( cut-off score or cut-off ) is understood to be the value determined by a test which, on the basis of a subjective or objective agreement, distinguishes between people with regard to an allocation (e.g. application of a diagnosis, need for treatment , Passing an examination, suitability for a job).

There are two different starting points:

  • Test separation value as a category limit: If there are differences in the characteristics of a characteristic within a group of characteristics, a test separation value can be used to categorize the characteristics. The determination of the test separation value is arbitrary and the category in this sense is artificial. For example, the line to giftedness is drawn with an IQ of 130, although the transition is fluid.
  • Indicator for a category: If two groups differ with regard to a characteristic, the test value of this characteristic can be used as an indicator of group membership. A clear example would be that the average height of women and men is different and therefore the measured height can be used as an indicator of gender.

In the first case one frequency distribution is separated, in the second case there are two frequency distributions.

Relationship with specificity and sensitivity

The determination of the test separation value is influenced by the misclassification rate to be accepted . Assuming that the target group to be found has higher values ​​in the test, the following applies: The higher the test separation value, the fewer people in the target group are recognized (low hit rate or sensitivity ). The lower the test separation value, the more people in the target group are recognized - but the higher the proportion of people who do not belong to the target group as a "false positive" assignment (low specificity ).

How high the test separation value is chosen depends on whether it would be more disadvantageous not to recognize people in the target group or to incorrectly assign people who do not belong to the target group.

There are four methods of setting a test cut-off value:

  • Two standard deviations above the test value of the healthy,
  • Two standard deviations below the mean of the sick,
  • Whether test subjects are closer to the mean of the healthy or the sick.
  • The ROC analysis can also be used to determine the threshold value . The costs of a misclassification are therefore essential for determining a cut-off value, whereby the costs for false-negative decisions can differ from those for false-positive decisions. Cut-off values ​​can be determined based on empirical data ( norms ) or theoretically (see above, criteria-oriented tests). The separation also depends on whether and how well the objects (people) to be separated differ in the corresponding feature and how high the measurement error of the test is.

See also

Individual evidence

  1. ↑ Katrin Rentzsch, Astrid Schütz: Psychological diagnostics: Basics and application perspectives . Kohlhammer, Stuttgart 2009, ISBN 978-3-17-019840-1 , pp. 24 ( limited preview in Google Book search).
  2. ↑ John W. Foreman, Jutta Schmidt: Smart data instead of big data: How you can get the best out of your customer data with Excel analyzes . Wiley, 2015, ISBN 978-3-527-69224-8 ( limited preview in Google Book Search).
  3. ↑ Axel M. Gressner, Torsten Arndt: Springer Lexikon Clinical Chemistry: Medical Laboratory diagnosis of A to Z . Springer, 2005, ISBN 978-3-540-23660-3 , pp. 1168 ( limited preview in Google Book Search).
  4. ^ Heinrich Tröster: Early detection in childhood and adolescence: strategies for developmental, learning and behavioral disorders . Hogrefe, 2009, ISBN 978-3-8409-2078-3 , pp. 109 ( limited preview in Google Book search).
  5. ^ A b Frank Schneider: Developments in Psychiatry: Symposium on the occasion of the 60th birthday of Henning Sass . Springer, 2006, ISBN 978-3-540-30100-4 , pp. 369 ( limited preview in Google Book search).
  6. ↑ DORSCH Lexicon of Psychology
  7. ^ Franzis Preckel, Miriam Vock: Giftedness: A textbook on fundamentals, diagnostics and funding opportunities . Hogrefe, 2013, ISBN 978-3-8409-2467-5 , pp. 104 ( limited preview in Google Book search).
  8. ↑ Markus Bühner : Introduction to the test and questionnaire construction . Pearson Deutschland GmbH, 2011, ISBN 978-3-86894-033-6 , p. 293 ( limited preview in Google Book search).
  9. ^ Christian Lenk, Gunnar Duttge, Heiner Fangerau: Handbook Ethics and Law of Research on Humans . Springer, 2014, ISBN 978-3-642-35099-3 , pp. 26 ( limited preview in Google Book search).
  10. ^ Christian FG Schendera: Regression analysis with SPSS . de Gruyter, 2014, ISBN 978-3-11-036252-7 , p. 174 ( limited preview in Google Book search).