A frequently used methodology for the analysis of different kinds of health problems is based on regression models: lineal, logistic, etc.; these models depend on a set of parameters that must be estimated from the data. However, they present the drawback of being very rigid since, occasionally, they impose overly strict relations between the variables. Non-parametric regression methods present the advantage of not establishing a priori restrictions, allowing the data to indicates us the appropriate functional form. In this paper several modern non-parametric regression methods are presented that in addition to their usefulness per se can prone to be of invaluable help in the diagnostic process for parametric regression methods. The current availability of the necessary software should contribute to their increased use which, in turn, will probably lead to an improved understanding of the health problems under study.