Multivariate Analysis

Abstract

In pathology and laboratory medicine, a common problem is to quantify and understand the contribution of multiple risks/hazards/exposures (or test results) to an outcome like disease status. Furthermore, we often want to combine multiple tests or clinical findings to create diagnostic/prognostic models. These goals can be achieved using multivariate analysis. The most common way to fit one (or more) variables to another variable and determine association or create predictive models is to create a linear regression model. Generalized linear models expand the concept of linear regression to continuous and non-continuous response variables like nominal or ordinal variables. There are different categories of generalized linear functions including linear regression, logistic regression, and Poisson regression.

In this chapter, we explain multiple regression analysis methods including logistic regression and ordinal regression. We also introduce the concept of statistical model validity and introduce the methodology for checking the validity of statistical models.