Simpson's Paradox refers to cases where a trend apparent in every subpopulation is reversed in the overall population. For instance, Treatment A could be performing better than Treatment B in treating large kidney stones as well as treating small kidney stones, yet when the data on treatment of both the type of kidney stones are aggregated, Treatment B could be found to be performing better. This is a potential problem for data analytics, and could lead to incorrect decision-making.

What are the main effective solutions for reducing the impact of Simpson's paradox in the data? Explain at least four of them.

Some of the references would be much appreciated.