deletions | additions       diff --git a/results.tex b/results.tex  index 3b4e978..eec26d5 100644  --- a/results.tex  +++ b/results.tex 
...
Figure ? shows comparisons of SPF vs SRIF attempting to fit simulated data to a given model. Note that for certain starting conditions, SPF outperforms SRIF for a large fraction of the total fitting time. SRIF pays an initial overhead cost due to the need to generate a new derivative matrix per iteration, as well as to perform the Householder operations. These overheads mean that each SRIF iteration is substantially longer than each SPF iteration. However, if the fit takes long enough, SRIF tends to converge before SPF (if SPF converges at all). Tables ? and ? show overall run statistics for the tests plotted in figure ?. Note that SRIF performed an order of magnitude faster than SPF for the one test that SPF converged.  \subsection{Real data: ET70}  if new shape has a better convergence (see comments): I have run \textbf{shape} with the new fitting routine on the asteroid 2000 ET70. ET70 has had its shape fit in the past, using \textbf{shape} and the SPF fitting routine (reference shantanu). \textbf{shape} with SRIF has converged on a lower $chi^2$. $\chi^2$.  as indicated in figure ? \par  if new shape has a worse convergence: I have run \textbf{shape} with the new fitting routine on the asteroid 2000 ET70. \textbf{shape} was run initially with an ellipsoid model. The starting conditions for the this model were such that the ellipsoid axes were all equal (Figure ?a). The best fit ellipsoid model (Figure ?b) was then converted into a spherical harmonics model with 200 model components. This spherical harmonic model was then fit as well (Figure ?c). The initial ellipsoid model fit resulted in a $\chi^2$ drop of $34\%$. The spherical harmonic fit resulted in a $\chi^2$ drop of $13\%$. Figure ? indicates the surface gravity field for ET70 given the final shape. Table ? gives the moments of inertia for ET70 given the final shape.  \par (the above was a clunky paragraph)