The "Least Square Criterion" requires which?

Values with the lowest values make the best predictors

The square of the regression solution is the best predictor

The lowest number of subjects that will provide a prediction is best

The square of error values must have their lowest possible value