Hi Andy, Andrew,
I was about to mention looking at the end of 'covar.txt'... Andrew
beat me to it.
Much of the rest of 'covar.txt' is basically just debugging data. I
see that it shows, for example, unit vectors along the velocity axis
("fwd"), perpendicular to that first vector and to the orbit plane
("up"), and perpendicular to both of these and therefore in the orbit
plane ("sid" as in "sideways"). Then you get some details from the
least-squares computation, followed by the covariance and correlation
matrices... which are in terms of the state vector (x, y, z, vx, vy, vz),
rather than in terms of orbital elements. Doing a least-squares fit in
terms of orbital elements has some drawbacks, such as dealing with
singularities when the inclination or eccentricity are near zero.
So internally, Find_Orb thinks "state vector" and converts to elements
when it's time to show things to mere humans.
Then you get eigenvectors of the correlation matrix, because I've
some hopes of using these to determine the line of variations (still
not working, %*&!@ it) and then the aforementioned sigmas.
Anyway, almost none of this is useful (at least not yet) except for
the sigmas. The rest of it is mostly useful only to me, when I'm trying
to figure out why Find_Orb is misbehaving in some area.
-- Bill
puck2017 wrote:
> Hi Bill,
>
> How would I go about decoding the covar.txt file if I all I wanted was the uncertainties in the 7 main orbital parameters?
>
> Thanks,
> Andy