Talk:Mean absolute scaled error

13 September 11:01 The equations for MASE and seasonal MASE are not correct. They are not the same as the equations in the cited source. The numerator and outer loop are problematic, since for some reason the in-sample error is used, and the outer loop is looping on the same variable as the inner loop, in a way that doesn't make sense. I will edit these when possible. — Preceding unsigned comment added by Sidneyradcliffe (talk • contribs) 09:05, 13 September 2018 (UTC)

This is not accurate, the original paper gives $q_{t}={\frac {e_{t}}{{\frac {1}{n-1}}\sum _{i=2}^{n}|Y_{i}-Y_{i-1}|}}$ and that the MASE= $mean|q_{t}|$. So the formula given in the wiki article is the correct one. — Preceding unsigned comment added by 160.39.16.142 (talk) 13:44, 5 January 2019 (UTC)

The equation you have put here, from the paper, is correct, whereas the one on the wiki page is not quite right. It is important to note that 'n' can be different from 'T'. 'n' is the number of 'in samples'. Also, the equation in the article assumes that you will be forecasting from t=1, whereas a common case is to forecast from t=n+1 to t=n+h, i.e. an out of sample forecast with a horizen of h. Basically the way the MASE is formulated in the paper is much much more preferable to the way it is formulated in this article. Sidneyradcliffe (talk) 10:50, 16 April 2019 (UTC)