My question is bit off from SAS and STAT tracks. But, hopefully you guys might have also come across this kind of puzzle.
I am handling few 3-way interactions and have applied Tukey’ s student range test and Tukey-Kramer’s test to significant differences among means. Y variables are log transformed. I want to take opinion what should be the best possible way to present these interactions. I am thinking to have it in tables instead of graphs. But adding three different schemes of letters/numbers/symbols to differentiate means are making tables congested. In my field people, mostly put LSD. But I want to stick to Tukey’ s tests.
Can we calculate something like LSD based on Tukey’ s test to get rid of different schemes of letters/numbers/symbols in the table? Are there any good reference showing 3-way interactions in neat form?
Although I am generally not a fan of PROC G3D, it has a SCATTER statement which can be used along with the SHAPE option to show the groups in the interaction. Code the various shapes to the significantly different groups in the interaction - this can work well when the groups are fairly distinct.