ABSTRACT
Multivariate modeling has increased the prevalence of statistical fallacies in biological inference. If two independent samples have been measured on the same two variables, a test to see if the two correlation coefficients are significantly different from each other is equivalent to a test to see if the two regression coefficients are significantly different. One of analysis of covariance's (ANCOVA) important assumptions is equality of regression coefficients across the populations that correspond to the study's comparison groups. It should be noted that different kinds of suppression can occur in multiple regression, depending on the sign and size of the bivariate correlations and the standardized regression coefficients. The term bouncing betas is sometimes used to describe the fact that a predictor variable's estimated beta weight is highly likely to change as other predictor variables are added to or deleted from the regression model.
