So, when we're playing with outliers, and then checking to see if everything has been taken care of, it quickly becomes obvious that , depending on what is left in and what is left out - which is totally the judgment of the researcher - one can get very different results. In fact, I'd probably guess that the whole process is, in a sense, path dependent; removing certain observations diagnosed as outliers leads to a new truncated data-set, which has its own anomalous observations, repeat ad infinitum. So, how what's our criterion for when to stop shaving, tinkering, and playing, and let the data speak for itself? It would seem that this threshold is very important considering the messiness of data in the social sciences. Thoughts? Andy _______________________________________________ gov1000-list mailing list gov1000-list(a)lists.fas.harvard.edu http://lists.fas.harvard.edu/mailman/listinfo/gov1000-list