It's an interesting question, and actually what the Advanced Output of Arena was designed for.
Let's assume a relatively simple setup:
-4 vertical LED tiles, each 192x384
-Placed next to each other in the processor, in a 1024x768 output.
-Placed next to each other in the real world, with a similar sized gap between them
-Our source content is in a 1920x1080 comp.
This means that we would need 7 identical horizontal tiles to recreate our real world setup ( 4 for the tiles, 3 for the gaps between them ).
This would mean our input slices have to be 1920/7 =~ 274 pixels wide. Since the tiles have 1x2 aspect ratio, they would need to be 274*2 = 548 pixels high.
Place these input tiles horizontally across your comp, leaving space for the gaps. Place them vertically centered, so the center of your comp stays the center of your output as well.
In the output, simply resize them to 192x384 (the size they have in the LED processor), and place them next to each other.
On an actual job, you're probably not going to have the same sized gaps as tiles, so the math becomes a little more complicated. In general it's the same principle though, and you can usually judge it by eye.