Comparison of treatment means: a statistical fantasy.

Over the last several decades, there has been a continuing debate between scientists and statisticians concerning appropriate techniques for elucidating the relationships among treatment means. To shed some light on this murky subject, the authors interviewed a scientist of world renown, Dr. Henri Hubris, Professor of Herpetology, Northeastern Central University. Dr. Hubris is widely revered among statisticians for his rigid adherence to the Highest Statistical Standards (HISS). We thought his examples and discussion were so illuminating that we decided to present them here, to assist o ther scientists in determining appropriate mean comparison procedures to use in their data analyses. Dr. Hubris began the interview by pointing out that treatment comparisons are "a real snake pit." The selection of appropriate statistical procedures for analyzing data depends on the relationships among the var ious expe r imen ta l t r ea tmen t s ( t r ea tmen t s t ructure) . T h e t r ea tmen t structure is a separate consideration from