Author (year)



“The use of the pseudo-orthogonal design biases the differences in means for the main effects relative to the differences in those means that would be obtained in a single-factor experiment” (p. 464).


“Dichotomizing one variable at the mean results in the reduction in variance accounted for to 0.647 r2; and dichotomizing both at the mean, to 0.405 r2” (p. 249).


“Analyses with categorized continuous variables required greater than 40% more patients for the same power as that achieved using continuous variables” (p. 138).


“Dichotomizing a continuous predictor variable can be conceptualized as adding an error of measurement to the variable. As a result, the effects of dichotomization are similar to the effects of random error of measurement” (p. 186).


“Dichotomization of continuous data is unnecessary for statistical analysis and in particular should not be applied to explanatory variables in regression models” (abstract).


“Dichotomizing a continuous variable is known to result in the loss of information, lower statistical power, and lower reliability” (abstract).


(Dichotomization) “(…) is harmful from the viewpoint of statistical estimation and hypothesis testing” (abstract).


“Modern regression models do not require categorization. In general, continuous variables should remain continuous in regression models designed to study the effects of the variable on the outcome of interest” (p. 3).


“Undesirable effects occur from dichotomization of both independent and dependent variables. The problem gets worse when multiple independent variables are split; for example, residual confounding is introduced, and spurious interaction effects may be seen” (p. 225)


“Simply dichotomizing continuous variables without previously referring to the original distributions by plotting them and checking consequences of dichotomization is a bad idea and should be discouraged” (p. 78).