Statistical techniques for possible identification on bivariate relations

Abstract

Introduction:

Most of the problems in biomedical research are of causal nature. The analysis of these studies should begin with the search for an association between the variables that represent the cause and the effect, and only if the association is significant will the causal inference analysis continue.

Objective:

To systematize the different statistical techniques that verify a bivariate relationship according to the type of variable.

Methods:

An exhaustive bibliographic review on the subject was carried out in the biomedical databases hosted in the Internet. The content was organized by sub-topics and a material with a critical synthesis of the most important aspects was elaborated, in which the experience of the authors was also expressed.

Results:

According to the type of variables, we have presented basic information about the coefficients, hypothesis tests, and graphs used in each case, the association measures to study risk, the features that ensure the validity of an association; chance and bias are also exposed as the mistakes that could be made in the investigation process and that could invalidate the existence of an association. The way of analyzing the association in the implicative statistical analysis is also presented.

Conclusions:

The knowledge of statisticians to verify a relationship between variables and the selection of statistical techniques is essential for carrying out the initial process of causal inference.

Keywords:   statistic techniques; bivariate relations; association; correlation; odd ratios; implicative statistical analysis.