Isn't this ruling out outlier observations given our model, rather than evaluating our model given outlier observations?