Burles, Morad, Bolton, and I discussed a project (Lexi Moustakas's idea) to find double LRGs in the SDSS LRG spectra by finding spectra with non-gaussian absorption line profiles. The basic procedure should be:
- Read in Schlegel's amplitudes for the Elodie spectral fits to the LRGs, along with his velocity dispersions, if all such exist in the spectro directory.
- Construct the gaussian-smoothed, linearly combined Elodie spectrum for each LRG.
- Difference the smoothed, combined Elodie spectra with others that have additional skewness and kurtosis (at fixed variance) to make "derivative" spectra that are derivatives with respect to skewness and kurtosis.
- Make "residual" spectra that are the differences between the LRG spectra and their gaussian-smoothed Elodie models.
- Compare the residual spectral with the derivative spectra and plot the "amplitudes" of the derivatives for all residual spectral vs each other and the velocity variances.
- Inspect outliers, which may or may not be double-LRGs (or other oddities).
We also discussed the reconstruction of full 2d (spatial) by 1d (spectral) spectral "cubes" from grism (or slitless) spectra taken at multiple position angles; we all concluded that it is possible. Burles had the good idea not to allow arbitrary spectra at every pixel, but to restrict spectra in each pixel to lie in some N-dimensional subspace of spectrum space.