The researchers looked at different ways to estimate a model that predicts outcomes based on multiple factors. They found that using a simpler method called maximum likelihood estimation works just as well as a more complex method called GMM estimation. They also found ways to improve the model by considering differences between the individuals in the data.