Chapter 5 - Logistic Regression

Stephen Roecker and Tom D’Avello

2020-03-13

Objectives

Pesky ‘Linear Model’ Assumptions (review)

Comparison between Linear Models and GLMs

Type Distributions Estimation Method Goodnesss of Fit
Linear Gaussian (i.e. Normal) least squares variance
GLM Any Exponential Family maximum-likelihood deviance

Generalized Linear Models (fewer assumptions)

Family (or Distribution) Default Link Function Data Type Example
Gaussian identity interval or ratio clay content
Binomial logit binary (yes/no) or binomial (proportions) presense of mollisols
Poisson log counts # of species

Overview - Logistic Regression

Example 1: Probability of Mollisols

Beaudette & O’Geen, 2009

Beaudette & O’Geen, 2009

Example 2: Probability of Red Clay

Evans & Hartemink, 2014

Evans & Hartemink, 2014

Example 3: Probability of Ponding