On 15-Jul-2006, at 17:14, George (Poonkhin) Khut wrote:
> The output of the standard deviation reports a
> strange value in the number box: "-1.#IND"
There are a couple of "strange" strings that are actually pretty
common: #INF, #NAN. I know what they are (calculations that result in
infinity or undefined values–’not a number’–respectively). #IND was
new to me, but Googling indicates that the C libraries used on
Windows report #IND for not-a-number im certain circumstances.
There a variety of explanations for this, all depending on the input
data, and they are not necessarily bugs.
When using lp.stacey in window mode, it is quite possible for
floating-point precision errors to accumulate. This is also an issue
with other objects that maintain running averages and the like. I
seem to recall average~ having a similar issue that was discussed a
few months back.
I will take a closer look at stacey’s code to see if I can handle
these problems in a way that is less surprising to the user. It would
be extremely helpful to have a data set that produces the error.