Finding local minima and maxima of a graph

[Bryan McCormick]

Jonathan, Jerry, et al

As to the number of potential observations. Thousands of series, 10s to 
hundreds of thousands of observations per series at a minimum. Possibly 
millions. So computational efficiency isn't a luxury, it is a requirement.

Yes, absolutely, detrending is required. I won't go into the details, 
but suffice to say every model I've seen so far advises this as a step 
to make sure one has significant observations. It will make real peaks 
and troughs much more apparent.

In stats packages, or even Excel, any of the above is trivial. 
Unfortunately in LiveCode there isn't (apparently) much of a base of 
functions to start with.

No, no forecasting. This is merely one step of gathering of data for 
additional observation.

Sure, a rolling window makes sense if you have to step through in a 
semi-brute force manner. If it comes down to it, I'll do it that way. It 
would seem to me there just has to be an easier way to do it.


On 12/5/2010 12:33 AM, Bryan McCormick wrote:
> Jonathan,
>
> Sure, that approach can work. But it isn't terribly efficient. It would
> however be less computationally intensive if I detrend first. That's the
> first step. So if you have a notion on a least squares linear fit line,
> that would be most welcome.