But if we have no data for test=[2], then that prior should be our best indicator of what the prediction should be, right? If not the model itself is the problem, correct?
But if we have no data for test=[2], then that prior should be our best indicator of what the prediction should be, right? If not the model itself is the problem, correct?