Standard Errors of Correlations Adjusted for Incidental Selection.

Autor/inn/en	Allen, Nancy L.; Dunbar, Stephen B.
Titel	Standard Errors of Correlations Adjusted for Incidental Selection.
Quelle	(1988), (22 Seiten)Verfügbarkeit
Sprache	englisch
Dokumenttyp	gedruckt; Monographie
Schlagwörter	Computer Simulation; Correlation; Error of Measurement; Matrices; Statistical Bias; Test Validity + Suchen Sie Ihr Suchwort? Computergrafik; Computersimulation; Korrelation; Messfehler; Matrizenrechnung; Testvalidität
Abstract	A recurring problem in educational research is how to account for non-random selection that has restricted the range of the variables of interest in correlational analyses. Several expressions due to H. Pearson (1903) and presented in matrix notation by D. N. Lawley (1943-44) are commonly used in selection settings to adjust for samples chosen on the basis of a direct selection variable. When validity evidence is provided for tests used to select applicants or for tests administered to examinees selected for some other reasons, correlations are often used. On the basis of this study, two suggestions can be made to practitioners using the Pearson-Lawley adjustments in these settings. The first is a recommendation for the use of the estimates for the standard error of the adjusted correlations. The second is a warning concerning the use of an adjusted correlation without reporting an estimate of its standard error. Since these suggestions are made in the context of an exact fit of the data to the Pearson-Lawley assumptions of linearity and homogeneity of variance, they indicate "best case" results for real data that may not fit the assumptions exactly. A computer simulation contributed new information about standard errors of adjusted correlations in three areas. Nine tables are included. (TJH)
Erfasst von	ERIC (Education Resources Information Center), Washington, DC