It is probably easiest to break this down into three problems:

• choose number of arrays x to draw from
• for each number of arrays, create the (7 choose x) combinations of arrays to use
• for each combination of arrays, iterate over all the elements in each array

The first two tasks could be combined by assigning an array to each bit of a binary number, and iterating over (15..127) and only accepting binary numbers with at least 4 ones in it. But this sort of strategy is not efficient for large numbers of arrays.

To iterate over the elements on each array, you can again use a counting scheme. For the selected arrays @a1,..@ak (k = 4,..,7), form the number N = @a1*...*@ak. Then count from z = 0 to N-1. This number z represents a unique combination of the element positions in your array.

• int (z/(@a2*...*@ak)) mod @a1 is the ele position in @a1
• int (z/(@a3*...*@ak)) mod @a2 is the ele position in @a2
• ...
• z mod @ak is the ele position in @ak

Sorry, don't have time to code it up.