hi David, in the notes for modelling and forecasting (Diebolds), you stated that;"it is clear that the model with least MSE also depict least sum of squared residuals owing to the fact that the division by T does not
have significant impact on the sorting of models done on the basis of their respective MSE outcome."
does this mean that when ranking models, we can just ignore the sample size (T) and focus on the squared differences between the observed variables and model's predicted variables of each model?
will it also be right to ignore squaring the difference and focus on absolute difference only, since squaring of the difference only doubles the numerator of the formulae and wont actually cause any error in terms of ranking models using only absolute differences ?
have significant impact on the sorting of models done on the basis of their respective MSE outcome."
does this mean that when ranking models, we can just ignore the sample size (T) and focus on the squared differences between the observed variables and model's predicted variables of each model?
will it also be right to ignore squaring the difference and focus on absolute difference only, since squaring of the difference only doubles the numerator of the formulae and wont actually cause any error in terms of ranking models using only absolute differences ?
