Using superimposed orthogonal Latin squares to construct a balanced tournament design, I get an n x n array of unique ordered pairs. If I only want unique unordered pairs, I can eliminate half of the "broken" diagonals (the diagonals other than the main diagonal), replacing their content with empty cells, to get a Room square or Howell design.
For example, in a square of 7 x 7, I find that the diagonals parallel to the main diagonal that start in columns 1, 2, and 4 contain the same set of pairs as the diagonals that start in columns 6, 5, and 3, respectively, but with the symbols reversed, (x,y) (y,x). I can eliminate the cells in either of those two sets of diagonals and preserve the property that each symbol appears only once in each row and in each column. Intuitively, I suspect that this pattern is a result of the fact that I transposed my Latin square to obtain its orthogonal Latin square.
In this example, I used (6i + 2j) mod 7 as my composition operator and obtained the following juxtaposed orthogonal Latin squares.
M | 0 1 2 3 4 5 6
-----+-----------------------------------
0 | 1,- 3,7 5,6 7,5 2,4 4,3 6,2
1 | 7,3 2,- 4,1 6,7 1,6 3,5 5,4
2 | 6,5 1,4 3,- 5,2 7,1 2,7 4,6
3 | 5,7 7,6 2,5 4,- 6,3 1,2 3,1
4 | 4,2 6,1 1,7 3,6 5,- 7,4 2,3
5 | 3,4 5,3 7,2 2,1 4,7 6,- 1,5
6 | 2,6 4,5 6,4 1,3 3,2 5,1 7,-
Eliminating diagonals 3, 5, and 6 yields a Room square. Alternatively, diagonals 1, 2, and 4 could have been eliminated.
M | 0 1 2 3 4 5 6
-----+-----------------------------------
0 | 1,- 3,7 5,6 2,4
1 | 2,- 4,1 6,7 3,5
2 | 3,- 5,2 7,1 4,6
3 | 5,7 4,- 6,3 1,2
4 | 6,1 5,- 7,4 2,3
5 | 3,4 7,2 6,- 1,5
6 | 2,6 4,5 1,3 7,-
For relatively small squares, it's not difficult to choose which diagonals to eliminate by inspection and a bit of trial-and-error. However, I'm not comfortable with that as a solution. Is there a construction that will directly show which cells to eliminate?
[This isn't a homework question; I'm not a student.]
The 2nd table is essentially a starter-adder construction of a Room square and so you need to choose the diagonals so that the pairs and their position in row zero meet the defined properties of starter and adder respectively. I believe the literature on starter-adders is going to give you more Room square constructions than you can get starting from pairs of orthogonal Latin squares.