Fitting to Conic Sections


Astronomy research
Software instruments
   Stellar equation of states
   EOS with ionization
   EOS for supernovae
   Chemical potentials
   Stellar atmospheres

   Voigt Function
   Jeans escape
   Polytropic stars
   Cold white dwarfs
   Adiabatic white dwarfs

   Cold neutron stars
   Stellar opacities
   Neutrino energy loss rates
   Ephemeris routines
   Fermi-Dirac functions

   Polyhedra volume
   Plane - cube intersection
   Coating an ellipsoid

   Nuclear reaction networks
   Nuclear statistical equilibrium
   Laminar deflagrations
   CJ detonations
   ZND detonations

   Fitting to conic sections
   Unusual linear algebra
   Derivatives on uneven grids
   Pentadiagonal solver
   Quadratics, Cubics, Quartics

   Supernova light curves
   Exact Riemann solutions
   1D PPM hydrodynamics
   Hydrodynamic test cases
   Galactic chemical evolution

   Universal two-body problem
   Circular and elliptical 3 body
   The pendulum


   Zingale's software
   Brown's dStar
   GR1D code
   Iliadis' STARLIB database
   Herwig's NuGRID
   Meyer's NetNuc
cococubed YouTube
Bicycle adventures
Public Outreach
Education materials
2023 ASU Solar Systems Astronomy
2023 ASU Energy in Everyday Life

AAS Journals
2023 AAS YouTube
2023 AAS Peer Review Workshops
2023 MESA VI
2023 MESA Marketplace
2023 MESA Classroom
2023 Neutrino Emission from Stars
2023 White Dwarfs & 12C(α,γ)16O
2022 Earendel, A Highly Magnified Star
2022 Black Hole Mass Spectrum
2022 MESA in Don't Look Up
2021 Bill Paxton, Tinsley Prize

Contact: F.X.Timmes
my one page vitae,
full vitae,
research statement, and
teaching statement.

Lots of tools exist for fitting a set a data to a straight line. How about fits to a general conic $$ a x^2 + b x y + c y^2 + d x + e y + f = 0 \ , \label{eq1} \tag{1} $$ perhaps with a constraint that enforces a specific conic section? References I found useful include Thomas & Chan 1989, Halir & Flusser 1998, Fitzgibbon et al 1999, and Harker et al 2008.

For direct (i.e., not iterative) least-squares fitting of data to an ellipse, fit_ellipse.f90.tbz will generate noisy data for an ellipse (an interesting problem itself!), fit the noisy data to equation (1) with the constraint $b^2 - 4 a c \lt 0$ that assures an ellipse, and report key geometric attributes such as the center coordinates, foci coordinates, lengths of the semi-major and semi-minor axes, and rotatation angle of the conic. This tool currently depends on the lapack routines dgetrf and dgetri to invert a 3x3 matrix and the lapack routine dgeev to obtain the eigensystem of a 3x3 matrix.


For direct, least-squares fitting of data to a hyperbola, fit_hyperbola.tbz will do the same as above, except with the constraint $b^2 - 4 a c \gt 0$ that assures a hyperbola.

Parabolas, defined by the constraint $b^2 - 4 a c = 0$ with zero to near machine precision, lie on the boundary surface between ellipses and hyperbolas. Two of the figures below show that $b^2 - 4 a c = 0$ is an elliptical cone with its vertex at the origin. All ellipses lie within the cone, hyperbolas lie outside the cone. Direct, least-squares fitting of data to a parabola is thus more involved than for ellipses or hyperbola. The tool fit_parabola.tbz will do the same exercises as its ellipse and hyperbola cousins above.

image image

For direct, least-squares fitting of data to a circle where the measure of fit goodness is an area and not a length, the tool fit_circle.tbz will do the same as its cousins above.


For iterative least-squares fitting of data to a non-specific conic, the tool fit_nonspecific_conic.tbz will do a traditional singular value decomposition on equation (1) ala e.g., Gander, Golub & Strebel 1994. Here is an example of noisy parabolic data with the returned best fit coefficients suggesting the geometery is an ellipse.

Please cite the relevant references if you publish a piece of work that use these codes, pieces of these codes, or modified versions of them. Offer co-authorship as appropriate.