We are investigating to what extent one can use a P(D) analysis to extract number counts of unclustered sources from maps of the far infrared background. Currently available such maps, and those expected to emerge in near future are dominated by confusion noise due to poor resolution. We simulate background maps with an underlying two slope model for N(S) and we find that in an experiment of FIRBACK type we can extract the high flux slope with an error of few percent while other parameters are not so well constrained. We find, however, that in a SIRTF type experiment all parameters of this N(S) model can be extracted with errors of only few percent.

We introduce a simple model for the formation of voids. In this model the underdensity of galaxies in voids is the product of two factors. The first arises from a gravitational expansion of the negative density perturbation. The second is due to biasing: galaxies are less likely to form in an underdense region. One feature of the model is an upper cutoff in void sizes. We calculate the volume filling factor of characteristic voids for different CDM models and find that our formation model points to $\Lambda$CDM models as preferred models of the power spectrum. A natural consequence of our model is that the underdensity of the dark matter inside voids is smaller than the galaxy underdensity.