if we have data on gender, and we use more categories than male, female, how can we analyze these data, beyond descriptively, if (as is often the case) there is a small number in the other categories? Including them in their own categories will usually lead to poor estimation but it doesn't seem right to exclude them either. Does anyone know of any guidelines or have any suggestions?
------------------------------
Melanie Bell, PhD
Professor of Biostatistics
Mel & Enid Zuckerman College of Public Health
Statistics Graduate Interdisciplinary Program
University of Arizona
------------------------------