Ken Willard made the first of these kinds of configurations. I believe it was published in RCM a long time ago. You might see if the plans are still available.

A reasonable starting place is to assume that the airplanes on the two sides act as a single wing. The airplane in back is angled nose down to simulate a horizontal tail. The airplane in front is set at a zero angle to simulate the nose of an airplane.

Make a small balsa glider to check out the CG location and incidence angles.