Masters Thesis:

A Comparison of Least-Squares and Bayesian Techniques to Radial Velocity Data Sets PDF or Postscript
(Note figure 1 is missing, but is shown below (on this web page).)

Publications: (link to ADS list)
  1. A Comparison of Least-Squares and Bayesian Techniques in Fitting the Orbits of Extrasolar Planets 2005AAS...207.6822D
  2. The N2K Consortium. I. A Hot Saturn Planet Orbiting HD 88133 2005ApJ...620..481F
  3. Fitting Astrometric Data With Markov Chain Monte Carlo: A Tool for Detecting Planetary Signals 2004AAS...20513505D
  4. Search for Terrestrial Planets with the Space Interferometry Mission 2004AAS...205.3907M

More than 160 extrasolar planet have been discovered orbiting nearby stars in our Milky Way Galaxy.  Due to observational biasing, most of these planets are Jupiter mass (1 MJupiter = 318 MEarth) and have rather short orbital periods.  The California planet search team detects the presence of planets indirectly, through the motion of the primary star.  As a planet (including the planets in our solar system) orbits its primary star it induces a small gravitational wobble in the star.  If this wobble is aligned along our line of sight (radially) then the light from the star is doppler shifted.  This doppler shift is measured as a shift of the stellar spectral lines, and the radial velocity of the wobbling star is calculated.  The radial velocity of the star in time is a sinusoidal variation described by the derivative of Kepler's Law of orbital motion.  Kepler's law is nonlinear and contains at least six free parameters (per planet): period, amplitude, eccentricity, argument of perihelion, time of periastron passage, and amplitude off-set.  To find the best fit orbital parameters for a given radial velocity curve a parameter estimation technique must be used.  For simple data a frequentist method, such as Levenberg-Marquardt, suffices to find the most likely orbital parameters, or maximum likelihood estimation (MLE).  For less than ideal data sets with low phase coverage, number of observations, or unknown number of planets a more robust parametric techniqueA Typical Orbit is employed.  My research is in developing and testing a Bayesian fitting method called Markov chain Monte Carlo (MCMC), which uses a Metropolis-Hastings algorithm to search parameter space.   A convergent Markov chain is proportional to the posterior probability density function (PDF) for each parameter.  For more details see my thesis in progress.  Here are some links to expert Bayesian astrostatisticians.
Eric Ford (UC Berekely)
Phil Gregory (U British Columbia)
Tom Loredo (Cornell U)
Bill Jefferys (U Texas)

Here are some links to other experts in the field of planets and planet formation.
Debra Fischer (SFSU)
John Chambers (Carnegie Institute)
Derek Richardson (U Maryland)
Greg Laughlin (UC Santa Cruz)
Doug Lin (UC Santa Cruz)

All my orbit fitting research is done in the IDL programming language.  For more information about my team at SFSU visit here.