In a study of the causes of bearing wear, a machine was run 24 times, with various loads (denoted x1), oil viscosities (x2), and ambient temperatures (x3). The wear, denoted y, was modeled as y = β0 + β1x1 + β2x2 + β3x3 + β4x1x2 + β5x1x3 + β6x2x3 + ε. When this model was fit to the data, the sum of squares for error was SSE = 9.37. Then the reduced model y = β0 + β1x1 + β2x2 + β3x3 was fit, and the sum of squares for error was SSE = 27.49. Is it reasonable to use the reduced model, rather than the model containing all the interactions, to predict wear? Explain.