In exam analyses with a small number of candidates and/or items, the validity of the psychometric properties and/or of candidate performances in terms of interpretation is at times not sufficiently considered. By means of bootstrapping, we therefore address the question of from what number of candidates/items the psychometric properties/performances can be seen as reliable.
Objective
The results should form the basis for a discussion of the minimum number of candidates/items from which interpretations in multiple choice exams are permissible.
First results
Presentation at the GMA 2016 in Bern:
Hofer, Rainer; Huwendiek, Sören (16th September 2016). [Psychometric properties of multiple choice exams depending on the number of candidates and items: From which sample sizes are the psychometric properties reliable?]: In: Jahrestagung GMA. Abstractband. Bern. 14.-17.09.2016 10.3205/16gma175
Literature reference: Efron B. Bootstrap methods: another look at the jackknife. Ann Statist. 1979;7:1-26 DOI: 10.1214/aos/1176344552
Team
Dr. phil. Rainer Hofer
David Sichau
Prof Dr. Dr. med. Sören Huwendiek, MME