animals=read.table("http://www.unt.edu/rss/class/mike/data/animals.txt", header=T)
library(MASS)

#Logistic Regression
  GLM.0= glm(decision ~ 1, family=binomial(logit), data=animals)
  summary(GLM.0)
  AIC(GLM.0, k = log(nobs(GLM.0))) # BIC

  GLM.1 <- glm(decision ~ idealism + relativism + gender + research, family=binomial(logit), data=animals)
  summary(GLM.1)
  confint(GLM.1, level=.95, type="LR")
  AIC(GLM.1, k = log(nobs(GLM.1))) # BIC indicates a significantly better model

  #First compare the two model BIC (lower better).  Here is a statistical test.
  anova(GLM.0,GLM.1, test="Chisq")
  predict(GLM.1, type="response")
  table(GLM.1$y,fitted(GLM.1)>.50) #Except for rounding, this will reproduce the table from SPSS
  #226/315 = 71.7% correct classification


#DFA on the same data
  dfa.mod <- lda(decision ~ idealism + relativism + gender + research, data=animals)
  dfa.mod
  predclass=predict(dfa.mod)$class
  table(decision,predclass)  #228/315 = 72.4% correct classification, slightly better than logreg