Multiple Regression

The following provides a brief introduction to linear models which are implemented in a wide variety of statistical software such as R, Stata, SPSS, and SAS. See http://biostat.mc.vanderbilt.edu/BioMod for more information.

Simple Case: One Predictor

• Simplest possible case is one predictor that behaves linearly
• Why should we expect this to fit real data?
  • Unusual for relationships to be linear

General Case: p Predictors

• Can allow for a number of predictors
• Predictors can be expanded into multiple columns to relax linearity assumption
  • polynomials
  • piecewise polynomials (SplineFunctions?)