²  Even if a test statistic for the overall model structure is statistically not significant, we are still unable to confirm that the model is correctly specified, since a non-significant statistic only implies that there is not enough power to reject the current model.