Skip to main content

Table 4 Variance components of model 1 (without adjustment for rater stringency)

From: Cutting costs of multiple mini-interviews – changes in reliability and efficiency of the Hamburg medical school admission test between two applications

  2009 2010
Effect Variance components (95% CI) % Variance components (95% CI) %
Cand Systematic differences in candidate performance 0.13 (0.09-0.20) 12.2 0.21 (0.16-0.27) 18.1
Stat Systematic differences in station difficulty 0.06 (0.02-0.19) 5.0 0.04 (0.01-0.13) 3.3
Rater Systematic differences in rater stringency 0.09 (0.05-0.14) 7.8 0.04 (0.03-0.07) 3.8
Cand*stat Differences in candidate performance between stations 0.30 (0.25-0.36) 27.6 0.39 (0.34-0.43) 33.3
Rater*cand (+ error) Differences in rater candidate taste, and residual variance 0.52 (0.48-0.56) 47.5 0.48 (0.45-0.51) 41.5
Total 1.09 100 1.16 100
  1. Rel = Vcand./(Vcand + Vstat/nstat + Vrater/nratings per cand + Vcand*stat/nstat + Vrater*cand/nratings per cand).
  2. 2009: Rel = Vcand/(Vcand + Vstat/12 + Vrater/28 + Vcand*stat/12 + Vrater*cand/28) = 0.72.
  3. 2010: Rel = Vcand/(Vcand + Vstat/9 + Vrater/19.5 + Vcand*stat/9 + Vrater*cand/19.5) = 0.74.