Abstract
Ratio variables, emphasizing correlated denominators, are constructed in three ways. Five ratio variable multiple regression situations are constructed, each with five predictor variables.One of these displays strong "spurious" effects resulting from the use of ratio variables, Using Monte Carlo techniques, one hundred samples for each of four sample sizes (i.e., N =25, 50, 100, 200) are drawn and six average statistics are computed for five additional regression situations created with multicollinearity
†An earlier version of this paper was presented ar thc ar~nual meeting of the American Educational Research Association in Boston, 1980 (April). The authors would like to thank Richard D. Warren for his early support in pursuing this area, Karl F. Schuessler for his idea to begin this research and Derek F. Stubbs and an anonymous reviewer for their helpful comments on the paper.
†An earlier version of this paper was presented ar thc ar~nual meeting of the American Educational Research Association in Boston, 1980 (April). The authors would like to thank Richard D. Warren for his early support in pursuing this area, Karl F. Schuessler for his idea to begin this research and Derek F. Stubbs and an anonymous reviewer for their helpful comments on the paper.
Notes
†An earlier version of this paper was presented ar thc ar~nual meeting of the American Educational Research Association in Boston, 1980 (April). The authors would like to thank Richard D. Warren for his early support in pursuing this area, Karl F. Schuessler for his idea to begin this research and Derek F. Stubbs and an anonymous reviewer for their helpful comments on the paper.