Previous: Quick Overview

Up: Zelig Commands

Next: Details

Examples

• Use the turnout data set included with Zelig to estimate a logit model of an individual's probability of voting as function of race and age. Simulate the predicted probability of voting for a white individual, with age held at its mean:

  > data(turnout)
  > z.out <- zelig(vote ~ race + age, model = "logit", data = turnout)
  > x.out <- setx(z.out, race = "white")
  > s.out <- sim(z.out, x = x.out)
  > summary(s.out)
  

• Compute a first difference and risk ratio, changing education from 12 to 16 years, with other variables held at their means in the data:

  > data(turnout)
  > z.out <- zelig(vote ~ race + educate, model = "logit", data = turnout)
  > x.low <- setx(z.out, educate = 12)
  > x.high <- setx(z.out, educate = 16)
  > s.out <- sim(z.out, x = x.low, x1 = x.high)
  > summary(s.out)                                     # Numerical summary.
  > plot(s.out)                                        # Graphical summary.
  

• Calculate expected values for every observation in your data set:

  > data(turnout)
  > z.out <- zelig(vote ~ race + educate, model = "logit", data = turnout)
  > x.out <- setx(z.out, fn = NULL)
  > s.out <- sim(z.out, x = x.out)
  > summary(s.out)
  

• Use five multiply imputed data sets from () in an ordered logit model:

  > data(immi1, immi2, immi3, immi4, immi5)
  > z.out <- zelig(as.factor(ipip) ~ wage1992 + prtyid + ideol,
                   model = "ologit",
                   data = mi(immi1, immi2, immi3, immi4, immi5))
  

• Use the nearest propensity score matching via MatchIt package, and then calculate the conditional average treatment effect of the job training program based on the linear regression model:

  > library(MatchIt)
  > data(lalonde)
  > m.out <- matchit(treat ~ re74 + re75 + educ + black + hispan + age, 
                     data = lalonde, method = "nearest")
  > m.data <- match.data(m.out)
  > z.out <- zelig(re78 ~ treat + distance + re74 + re75 + educ + black +
                   hispan + age, data = m.data, model = "ls")
  > x.out0 <- setx(z.out, fn = NULL, treat = 0)
  > x.out1 <- setx(z.out, fn = NULL, treat = 1)
  > s.out <- sim(z.out, x=x.out0, x1=x.out1)
  > summary(s.out)
  

• Validate the fitted model using the leave-one-out cross validation procedure and calculating the average squared prediction error via boot package. For example:

  > library(boot)
  > data(turnout)
  > z.out <- zelig(vote ~ race + educate, model = "logit", data = turnout)
  > cv.out <- cv.glm(z.out, data = turnout)
  > print(cv.out$delta)