To calculate the mean squared prediction error of a model in test data, we take the mean of (predictions - observed outcome)^2.
Like the MSE, MSPE can be broken down
