Regression over regression: Or fixations about regression analysis

Abstract

Some of the problems in dealing with regression analysis are described from the psychostatistics analytical viewpoint. A case study and a brief history antecede the covering of these fixations of regression: stepwise is unreliable; with enough independent variables, R2 becomes 1; if R2 is less than .5, then it is too small to discuss; but you had a control group; but you did not have a control group; a regression analysis is not as sophisticated as ANOVA. The conclusion is made that sometimes the treatment of choice for these problems is to become a psychostatisical analysand of one trained in multivariate analysis.