Latest Releases

Installation instructions: http://projects.iq.harvard.edu/zelig/install

Version 4.2.2 (as Beta)

Available end of September

  • Weights, and robust standard errors enabled for more models.
  • Patch for plotting in "poisson.gee" model

Version 4.2.1 (on CRAN)

Released September 2013 as a feature update 

  • Improved summary statistics for multiply imputed data
  • Improved plots for multiply imputed data
  • Improved capability and flexibility in "plot.ci" function to graph quantities of interest over a range of covariates
  • Backwards compatibility in "plot.ci" calls to previous style, increased graphics parameter tunability (standard R "par" options enabled)
  • "plot.ci" now plots multiple confidence interval plots when both an "x" and "x1" parameter are passed to its "sim" object
  • "plot.ci" now supports NULL values for its "var" parameters for better backwards compatibility with previous argument calls
  • additional exemplar R "demos" enabled

    Version 4.1-3

    Released in February 2013 as a feature update:

    • Implemented Rubin's rule for use with imputed data
    • Adding the "subset" parameter to the "summary" function for Zelig calls that use multiple imputation
    • Adding additional parameters - main, sub, xlab, ylab, xlim, ylim -  to the "plot.ci" function. For more information, type help("plot.ci") from within an R-session
    • Updated installation instructions within documentation
    • Patches mislabled confidence interval plots.

    Version 4.1-2

    Released in January 2013 is a bug-fix update:

    • Restored Amelia compatibility to Zelig
    • Allows the use of "plot.ci" with imputed objects containing a single data set
    • Adjusted the display of "call" objects created when fitting statistical models
    • Changed how Statistical models are called internally
    • Updated describe methods for glm models
    • Fixed a bug where the "factor.bayes" model did not run correctly
    • Fixed a bug where the "by" parameter caused an error
    • Fixed a bug affecting statistical models using a single dependent variable
    • Fixed a bug preventing the simulation of the "relogit" model