data reliability estimation method, panel survey question selection/attenuation correction
Samenvatting
An analysis of a method for estimating reliability of data, requiring three subsequent measurements per individual on two measurements & a suitable instrumental variable. As examples, the selection of reliable questions in panel surveys & the correction for attenuation are discussed. For three-wave multiple-indicator data, the model for three measurements by David E. Wiley & James A. Wiley (see SA 18:7/70E5517) can be compared with L. J. Cronbach's alpha model ("Coefficient Alpha and the Internal Structure of Tests," Psychometrika, 1951, 16, 297-334). When using two real-life data sets, the former model produced slightly higher reliabilities, but when applied to two measurements & an instrumental variable, it proved less trustworthy. A method is proposed to select a suitable instrumental variable. 6 Tables, 4 Figures, 2 Appendixes, 28 References. Modified HA.
