The examples above are simple examples which only skim the surface of
R's plotting potential. We include more advanced, model-specific
plots in the Zelig demo scripts, and have created functions that
automate some of these plots, including:
- Ternary Diagrams describe the predicted probability of a
categorical dependent variable that has three observed outcomes.
You may choose to use this plot with the multinomial logit, the
ordinal logit, or the ordinal probit models (Katz and King,
1999). See Section for the sample code, type
demo(mlogit) at the R prompt to run the example, and refer to
Section to add points to a ternary diagram.
41#41
- ROC Plots summarize how well models for binary dependent
variables (logit, probit, and relogit) fit the data. The ROC plot
evaluates the fraction of 0's and 1's correctly predicted for every
possible threshold value at which the continuous Prob42#42
may
be realized as a dichotomous prediction. The closer the ROC curve
is to the upper right corner of the plot, the better the fit of the
model specification (King and Zeng, 2002b).
See Section for the sample code, and type demo(roc) at the
R prompt to run the example.
43#43
- Vertical Confidence Intervals may be used for almost any
model, and describe simulated confidence intervals for any quantity
of interest while allowing one of the explanatory variables to vary
over a given range of values (King, Tomz and Wittenberg,
2000). Type demo(vertci) at the R prompt to
run the example, and help.zelig(plot.ci) for the manual page.
Gary King
2011-11-29