[Therion] Aven colour by loop error

[Olly Betts]

On Sat, Apr 01, 2017 at 09:10:02PM +1300, Bruce Mutton via Therion wrote:
> The new capability of exporting loop closure information to 3d format is
> helping to identify the bad loops on our surveys.
> 
> Very nice.
> 
> However the scalebar at the top right seems to be locked into a range of 0.0
> to 12.0 (%) error.

Aven's error scale isn't in %, but rather in "sigmas" - i.e. it's how many
times the expected error the observed error is.  Assuming normal distribution
of errors (which sums of random errors will tend towards) anything more than 
3 is only 0.3% likely by random errors, so highly suspect:

https://en.wikipedia.org/wiki/68%E2%80%9395%E2%80%9399.7_rule

If what Stacho said back in February is still true, therion outputs the
wrong data in this field in the 3d file, so you actually see "2 * relative
error of a loop" there, by which I think he means percentage misclosure.

But as I said at the time, percentage misclosure isn't really a useful
measure of loop quality because the threshold of what is reasonable
varies with the length of the loop:

https://mailman.speleo.sk/pipermail/therion/2017-February/006300.html

> This is OK for our really nasty loop closures; the maps look suitably
> colourful, but some of our more recent cave surveys don't have error much
> more than 1.5%, so the picture is kind of shades of blue, as below.  It
> doesn't help with visualising the relative quality of the loops.

Unless Stacho has since fixed this, you're actually seeing 0-6% there
currently.

> Would it be possible to change the default scalebar to something like 0.0 to
> 2.0, with the upper limit stepping upwards to suit the data, if there are
> larger loop errors?

I'd much rather we fixed therion to export the correct error information.
Trying to colour based on percentage error just seems to be fundamentally
a less helpful approach.

Also, not varying the colours for a particular number of sigmas depending on
how bad the data is was a deliberate choice.  2 sigmas is no better or worse
just because someone blundered a survey elsewhere.  And if data is surveyed to
a lower quality the grades should be set appropriately to reflect that.

Cheers,
    Olly