Do different measures of assessment performance bias or exaggerate relationships?

Putwain, Dave (2007) Do different measures of assessment performance bias or exaggerate relationships? BERA Annual Conference, 5-8 September, Institute of Education, University of London.

Item not available from this archive.

Abstract

KS4 assessment performance data is used in at least three ways in educational research. First, an indicator of overall school performance to evaluate policy initiatives/ school practices designed to raise achievement. Second, to establish differential student achievement or mobility across socio-demographic groups (e.g. gender, ethnic background, etc.). Third, to establish how psycho-educational constructs such as achievement motivation, academic self-concept and assessment anxiety may influence students’ attainment. A standard convention for educational research in the UK is convert GCSE grades into an eight-point scale (e.g. Schagen & Schagen, 2005). There is less consistency over the calculation of performance measures and six different conventions are in current usage; total sum of points, total sum of points in the eight highest scoring subjects, grade point average (GPA), the mode of highest grades achieved, the number of five A*-C grades achieved and a binomial categorisation of whether a student achieved 5 A*-C grades or not. There is a possibility that particular measures may bias or exaggerate certain relationships. In order to investigate this possibility, an existing dataset regarding the relationship between GCSE performance and self-reported assessment anxiety was reanalysed using these six measures in a series of simple regression models. This reanalysis suggested that two measures, GPA and whether a student has achieved five A*-C grades or not, produce a stronger effect than the other measures. For example, GPA indicates a highly anxious student would achieve one grade lower across all subjects, whereas the total sum of points suggests a highly anxious student would drop three grades overall. These findings indicate that individual performance measures should be treated with some degree of scepticism and that in order to establish an educationally meaningful effect, multiple measures of performance should be triangulated

Item Type: Conference or Workshop Item (Paper)
Subjects: B Philosophy. Psychology. Religion > BF Psychology
L Education > L Education (General)
Divisions: Psychology
Date Deposited: 28 Feb 2011 13:46
URI: http://repository.edgehill.ac.uk/id/eprint/2463

Archive staff only

Item control page Item control page