Back to Search Start Over

Curvilinear Extensions to Johnson-Neyman Regions of Significance and Some Applications to Educational Research.

Authors :
Texas Univ., Austin. Research and Development Center for Teacher Education.
Wunderlich, Kenneth W.
Borich, Gary D.
Publication Year :
1974

Abstract

Considerable thought, research, and concern has been expanded in an effort to determine whether the assumption of a quadratic relation between a single predictor and a criterion violated the assumptions which Johnson and Neyman (1936) state for calculating regions of significance about interacting regressions. In particular, there has been special concern for the assumption of linearity. Debate has ranged from whether linearity referred to the functional relation of the criterion and predictor to whether it was in the context of a linear statistical model, if not both. This paper extends the Johnson-Neyman technique to the curvilinear case and then illustrates this extension by reanalyzing data from two previously published research studies which have considered only a linear relationship between a covariable criterion. A computer program is included in the appendix. (Author/RC)

Details

Database :
ERIC
Publication Type :
Conference
Accession number :
ED095221
Document Type :
Speeches/Meeting Papers