On interpreting results of multiple regression procedures: A cautionary note for researchers and reviewers

There appears to be increasing use of multiple regression procedures in clinically related research. The present paper discusses two problems regarding inappropriate use of, or reporting of results from, such data analyses. In particular, researchers and reviewers are alerted to the needs to partial the overlapping effects of predictor variables from one another in a conceptually meaningful order, and to report F values and their probabilities for the effects of individual independent variables, when interpretations are to be made about such effects, rather than reporting the F value for the entire equation at each step. Care also needs to be taken to correctly report the degrees of freedom of the F ratio. Two recent articles from this journal are used to illustrate these points. © 1987 Plenum Publishing Corporation.

Duke Scholars

Author Clive Justin Robins Psychiatry & Behavioral Sciences, Adult Psychiatry & Psychol ...