answer:I don’t have enough time right now to work out a solution, but here is what I quickly figured out: There 28 groups of three students and 4 groups of four students. Assuming the absolute worst-case scenario, you have reports from at least 16 of the 32 groups, so you are guaranteed to cut your work-load in half. It is probably easier to run a computer simulation. This problem is complex because not every group has an equal opportunity of being chosen, and the size of the sample doesn’t match up with the number of groups. If you choose a group of three, the probability of the group not being represented is slightly less than 1/8, but this isn’t very useful because the groups are not mathematically independent. I think that I can have the results of a computer simulation in a few hours.