Effects of Sampling on Measuring Galaxy Count Probabilities

We investigate in detail the effects of sampling on our ability to accurately reconstruct the distribution of galaxies from galaxy surveys. We use a simple probability theory approach, Bayesian classifier theory and Bayesian transition probabilities. We find the best Bayesian estimator for the case of low sampling rates, and show that even in the optimal case certain higher order characteristics of the distribution are irretrievably washed out by sparse sampling: we illustrate this by a simple model for cluster selection. We show that even choosing an optimal threshold, there are nonzero numbers for both misidentified clusters and true clusters missed. The introduction of sampling has an effect on the distribution function that is similar to convolution. Deconvolution is possible and given in the paper, although it might become unstable as sampling rates become low. These findings have important consequences on planning and strategies of future galaxy surveys.