|
PureBytes Links
Trading Reference Links
|
Gary,
> The StandardError function in TS2k suffers from the same error as many
> of TS's stat functions: they do subtractions of two large
> numbers to get a small result, which combined with TS's single-
> precision math results in horrifically inaccurate results or
> divisions by zero. Obviously TRAD's programmers never took a
> basic computational-accuracy course.
I always wondered why the calculations are single-precision. That always
seemed stupid to me.
> Once upon a time I rewrote the StdDev function to avoid this
> problem, but my attempts to do so for StandardError are only
> somewhat successful.
Isn't standard error exactly the same as the standard deviation of the
distances of values from a regression line? Then you could just calculate
the differences, and use your own version of StdDev to solve the standard
error.
-Alex
|