Are there any future plans for being able to specify a numeric variable in the PANELBY statement without having it use the individual values? The PROC would create groups out of intervals of the variable's values, similar maybe to Cleveland's slicing in "Visualizing Data".
Cleveland's approach to slicing interval data used his "equal interval" algorithm, which attempted to make the number of data points the same in each cell. The result was that the interval ranges could change per cell. In addition, he had a parameter that would allow you to specify overlap in the intervals. How much of this functionality do you need for your purposes?
As a side note, you could possible create this effect with the current version of SGPANEL by applying the algorithm to your data to determine the intervals and create a user-defined format for the interval ranges. SGPANEL will use the format to bucket your class data correctly for each cell.
Thanks for the response. Your side note is actually how I do it. I've been working on a macro of the algorithm. I was going to present it and a couple of other concepts (banking and cell ordering) at SGF, but I'm over my limit. :-(
The question comes up fairly often. I like the approach Cleveland took, though there could be better ones. I haven't looked in a while. Both the "equal-count" and the overlap make good sense for understanding the change in that dimension.
I wanted to see if there was direction already and, if there was, mention it in the paper I am presenting. (basic SGPANEL)