I've had an experiment in which random assignment failed and there are a priori differences in pretest.  

Or, a education/psychology experiment in which data with a priori pretest differences were published?  

Similar questions and discussions