Regression analysis for continuous independent variables in medical research: statistical standard and guideline of Life Cycle Committee

Abstract

Variables measured in medicine are usually continuous, such as blood pressure, serum glucose, and drug dosage. There are statistical methods to confirm whether these variables act as independent variables that cause an outcome. Most of these methods are types of regression analysis. In this article, starting with a simple correlation coefficient, linear regression, multi-factor regression analysis, analysis of covariance (ANCOVA), logistic regression, and modified Poisson regression will be explained. This article shall also cover when and how to use these regression analysis techniques. The following explanations can be used as a statistical background of research using regression analysis. It can also promote the understanding of medical researchers who would like to conduct advanced regression analysis.