The use of test scores from large-scale assessment surveys: psychometric and statistical considerations

Large-scale Assessments in Education

An IERI – International Educational Research Institute Journal

Table 1 Estimated regression coefficients for a model predicting wages (raw) based on skill variable theta (either true value, or PVs, or WLE, MLE, EAP estimates), gender and education

Estimator for θ	Intercept	COG	SEX	EDU
True ability	9.967	1.012	0.002	0.024
PVs	9.951	0.998	0.009	0.057
EAP	9.718	*0.810*	0.009	*0.539*
MLE	9.567	*0.548*	0.017	*0.871*
WLE	9.697	*0.668*	0.010	*0.592*

Simulation setup similar to the one found in von Davier et al. (2009)
The quantities printed in bolditalics and italics are those regression coefficient estimates that are adversely affected by the use of EAP, MLE as well as WLE estimates in the regression. These quantities are biased, while the estimates obtained using PVs are very close to the estimates calculated using the true person parameters