The researchers explored different ways to assign probabilities in statistics, focusing on a concept called the scale invariant prior. They found that by breaking down the total variance into smaller parts, they could create more flexible and adaptable models. This approach allows for a better understanding of how different factors contribute to overall variability in data.