Trend Vs LSL

[Sam Rowe]

Hi Guys,

Over the last week or so when we've had spare time, Happy and I have
been adding TREND to Jart. It's all working fine, but it has
limitations about how much future it can predict. We were using a
graph from the RRD Gallery (
http://oss.oetiker.ch/rrdtool/gallery/index.en.html search for trend )
as an example. Then I noticed that the example's TREND lines looked a
lot different than ours did. Ours are more average-y and his are more
MAX-y. Then I noticed that he's using Least Squares Line and not
TREND, so I quickly hacked in a version to use LSL and it looks much
more similar to the gallery image.

While we could offer both, I'd really rather offer the more useful one
and dump the other mostly because I don't want the UI to become even
more of a cluttered mess than it already is.

I can provide some example graph output if necessary.

What do you guys think?

[Mark Plaksin]

> While we could offer both, I'd really rather offer the more useful one
> and dump the other mostly because I don't want the UI to become even
> more of a cluttered mess than it already is.
>
> I can provide some example graph output if necessary.

Ya, we could really use guidance here. LSL sure *looks* like the best
choice. If we use LSL on the graph of our online learning database size
it predicts the steep slope of growth that we expect. If we use TREND
on that same graph it's obvious that the averaging is lowering the slope
of the future growth line.