I attempted to use glmD from the Design library and then robcov to get robust clustered errors from a probit model, but see now that Design is not able to do this because "robcov needs the residuals method for the fitter to allow a type="score" or type="hscore" (for Efron's method) argument. Until someone adds score residuals to residuals.glm robcov will not work for you" (http://tolstoy.newcastle.edu.au/R/help/04/06/0273.html). But that was 6 years ago, does anyone know if there is a canned program available now to do this?