cosmic ray cluster, LRG pairs, jackknife errors

I estimated specific star formation rates for the galaxies in the sky near the center of Farrar's putative cosmic ray cluster. We find no excess of star formation in that direction.

Morad and I discussed Richard Cool's (Arizona) testing of the SDSS photometry of close LRG–LRG pairs.

Morad and I figured out the following, tentatively: In what Eisenstein calls "jackknife" (I disagree, but that is a subject to discuss on Hogg's Universe, not here), you use your one data set to make N measurements of a quantity Q, for each of which you leave out (ie, do not use) a different 1/N part of your data. You take the root-variance of these N measurements about the mean, and multiply by sqrt(N-1) to estimate the uncertainty on your full-up measurement.


  1. In a game of rock-paper-scissors, bootstrap beats jackknife.

  2. Eisenstein hated the term "jackknife" also, but over the years it stuck. He also hated being called "The Velvet Fog."