Uncertainties warning

[ElChristou]

Guys, in many aspect we have to deal with uncertainties, from textures (lok) to star classes, binary orbits, uncertain DSO orientation, etc...

What about having an "Uncertainties tag" to place in a data file with a brief description of the problem. Approaching whatever element with such tag in it's declaration, would make Celestia to display a little sign somewhere on screen (probably under the data in top left corner). If the user click this sign, the description would appear next to it. If many uncertainties are in the observer's neighborhood, clicking this sign would list the descriptions.

(Note that many data files already have such notes about uncertain data in form of comments...)

Opinion?

[ajtribick]

This in a way also brings up the issue of the stars.dat format: given what is measured is RA, Dec and parallax, converting this into Cartesian uncertainties is an interesting task.

I've always been somewhat uneasy about using Cartesians in stars.dat. What was the reason for going to Cartesians again?

[selden]

Chaos,

One problem is that the math libraries on the different platforms give slightly different results. As a result, converting the contents of stars.dat from equatorial to cartesian (as is needed by Celestia internally) at load time yielded different positions on different platforms. Converting Stars.dat from equatorial to cartesian on one specific platform before distribution guarantees that those stars are at the same positions on all platforms.

Cel:: URLs are recorded in cartesian coordinates. URLs recorded on one system would place the observer at the same position on all platforms, but at different positions relative to their stars. I'm not sure why there wouldn't be similar problems for stars defined in STC files. I think those conversions are done at a higher precision than is used by the stars in stars.dat. (double precision vs single precision, perhaps?)

[t00fri]

Guys, in many aspect we have to deal with uncertainties, from textures (lok) to star classes, binary orbits, uncertain DSO orientation, etc...

Christophe,

I don't know why you chose the above examples of binary orbits and DSO orientation. The ones I included in Celestia are VERY accurate! We talk about a few percent in the orbits, for example. Of course if you get close enough, even 5% looks "big".

Everyone must be aware that that predictions based on ANY kind of scientific measurements NECESSARILY include a certain amount of uncertainty. The question is only HOW LARGE they are. That cannot be indicated with a simple tag...

Opinion?

No good in my view.

We must rather contemplate about a rendering trick to display actual uncertainties continuously. E.g. a "soft" orbit display. But such things are expensive, if done well...

Bye Fridger

[Chuft-Captain]

We must rather contemplate about a rendering trick to display actual uncertainties continuously. E.g. a "soft" orbit display. But such things are expensive, if done well...

Bye Fridger

I agree with this. Statistical chart data will often display the +/- uncertainties as bars/lines to either side above or below the data-point.
We need a more sophisticated (3D) version of this principle IMO.
Of course, it should be able to be switched on / off, and ideally configurable for different types of object.

(Not expecting this feature in 1.5.0 final )

[ElChristou]

Christophe,

I don't know why you chose the above examples of binary orbits and DSO orientation. The ones I included in Celestia are VERY accurate! We talk about a few percent in the orbits, for example. Of course if you get close enough, even 5% looks "big".

I'm not thinking in what we have but rather in what we don't have because of missing data (uncertainties too big)...

Everyone must be aware that that predictions based on ANY kind of scientific measurements NECESSARILY include a certain amount of uncertainty. The question is only HOW LARGE they are. That cannot be indicated with a simple tag...

Opinion?

No good in my view.

We must rather contemplate about a rendering trick to display actual uncertainties continuously. E.g. a "soft" orbit display. But such things are expensive, if done well...

Ok, but as you point, some case won't be easy to deal with. Here my point of view is general and at least a warning could cover easily whatever problem we can find and give us the time (if wanted) to do something more sophisticated for certain cases...

[samsmartguy]

I think it wouldn’t be bad to show uncertainties as bars/lines.