Researchers used a method called expectation maximization to estimate parameters of a K-distribution more accurately than other techniques. This method is better for large datasets and is more computationally efficient than traditional approaches.