The optimal window size for a WMA is a tricky one: too big and the WMA lags behind the data and does not respond quickly. Too small, and the WMA is very noisy. I suggest you start with 4 weeks window and visually examine the WMA vs. the actual data, if it is too noise increase the window size.
Another approach is to compute the root-mean squared errors (RMSE) between the WMA and original data, and try to find the data that gives the smallest RMSE value. RMSE is a function in NumXL.
评论
The optimal window size for a WMA is a tricky one: too big and the WMA lags behind the data and does not respond quickly. Too small, and the WMA is very noisy. I suggest you start with 4 weeks window and visually examine the WMA vs. the actual data, if it is too noise increase the window size.
Another approach is to compute the root-mean squared errors (RMSE) between the WMA and original data, and try to find the data that gives the smallest RMSE value. RMSE is a function in NumXL.
请先登录再写评论。