Mike,
Thank you for your kind reminder,there is a true numeric instability in the */\.

   (0.1&EMA ,&(>./) 0.1&EMA2)  i.6654
6644 6644
   (0.1&EMA ,&(>./) 0.1&EMA2)  i.6655
6645 _

So we must use it with caution when applying to stock market indexes.

ps,just try a c version,and the timing is 0.023s.


2008/3/11, Mike Powell <[EMAIL PROTECTED]>:
> Xu,
>
>  You need to be a little careful if you use the EMA2 algorithm in
>  production situations. There is the ever present problem of machine
>  precision. For example:
>
>  (0.1&EMA -: 0.1&EMA2) i.100
>  1
>     (0.1&EMA -: 0.1&EMA2) i.1000
>  1
>     (0.1&EMA -: 0.1&EMA2) i.10000
>  0
>
>  Usually, where economists and forecasters dwell, they are looking at
>  20 years and precision is not usually an issue. Mais, guardez.
>
>
>  Mike Powell
>  ----------------------------------------------------------------------
>  For information about J forums see http://www.jsoftware.com/forums.htm
>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to