Means separation tests on interaction effects can be difficult to interpret, due to the large number of means that may be involved. One way to simplify interpretation is to only consider mean separation within levels of one factor. This is called "slicing" by many people.
Suppose I have Irrigation and Spacing treatments. Default output might look like this:
But to assist interpretation, I want to focus on comparing Spacing levels within each Irrigation. For example, if I see that Spacing differences change across the Irrigation levels, that will help interpret any interaction between the two treatments.
"Slice" the interaction by using the SLICE= option in %MMAOV
and the resulting output is:
It now may be easier to see that Spacing=24 switches from overlapping with the highest group under No irrigation, to being completely in the lowest group with Yes irrigation.