[Regression and correlation]

Simple linear regression and correlation are frequently used in medical publications. The methods are overused and have become greatly confused, probably because of the close similarity between the mathematical calculations. Simple regression is a method for a mathematical description of the dependence of a random response variable of another variable, a regressor, which is not random. The model can be used to predict the response form given values of the regressor. Correlation is a measure of the degree of linear association between two random variables, and has nothing to do with regression. The method is mainly an investigative tool in a statistical analysis, or to suggest further research; for forming hypotheses rather than for testing them. To judge form publications, fitting of a regression line and calculation of the correlation coefficient is the method of choice in modelling the linear relationship between two random variables. This is a completely misguided approach. When both variables are random, least squares estimation, as used in regression, is not appropriate. The correct method is to estimate a linear structural model by maximum likelihood.