a guide to filtering the catalog

Post date: Oct 14, 2013 10:44:25 AM

We have developed a philosophy about this void catalog in which we attempt to report as many voids as possible and let you, the user, apply cuts and filters as you see appropriate. We have a set of filters that we apply for our own analysis, and for historical reasons this has become the "default" catalog. We also provide the full unfiltered catalog (except for a hard minimum size cut, to be discussed below) so you can create your own version of the catalog.

Some cuts you may consider applying:

- Size: We only provide voids with effective radii greater than the mean particle separation. However, the smallest voids are likely highly unreliable. Thus, you may want to only take voids with two times the minimum radius. You can find the radius in the "centers_" files.

- Minimum density: The watershed algorithm reports all basins, even if they have densities higher than you may like. You can, for example, only take voids with minimum Voronoi cell density < 0.2 or 0.3 times the mean density (as in Nadathur et al. 2013). Since minimum density scales inversely with void size, this also effectively removes most small voids. However, this quantity is very noisy, so your cuts may be more/less aggressive than you think. The minimum density can be found in the "voidDesc_" files. Since this is a raw ZOBOV output, it is expressed in units normalized to the total number of particles, including boundary particles. So you can either a) correct for the presence of boundary particles by rescaling by a factor of (n_g/n_tot)^3, or b) just look at the distribution of minimum densities and make a cut as you see fit. With the current version of the catalog n_g (number of galaxies) and n_tot (number of galaxies + boundary particles) are a bit difficult (but not impossible) to extract. We will provide these numbers more clearly in the next catalog release.

- Central density: This is similar in spirit to the minimum density above, but is typically computed in spheres about the barycenter. For the default catalog, we take spheres of 1/4 the void effective radius. This tends to affect larger voids, and becomes less effective for smaller voids (since we start to probe below the mean particle separation). If you want to calculate this on your own, then you have to use the dumpVoidParticles tool to extract the member particles.

- Density contrast: This is the ratio between the density on the void ridgeline and the minimum density. You can compare this to, say, the probability of getting that density ratio in a Poisson distribution (Neyrinck 2008), but this procedure is not necessarily appropriate for sparse, biased populations. You can find this quantity in both the voidDesc_ and centers_ files.

- Tree level: There is a full hierarchy of voids, subvoids, sub-subvoids, etc.. You may wish to take only top-level "parent" voids, or only leaf-level "children" voids. The default catalog contains only parent voids. You can find the tree level in the centers_ files.

Our default catalog contains only top-level parent voids with central densities < 0.2 times the mean density. If you are happy with these cuts, then you can just use the un-prefixed catalog.

If you want to make your own cuts, then start with the "untrimmed_" and help yourself.

In addition, we also have the "all" and "central" catalog division.

We would like to thank Seshadri Nadathur; our conversations with him prompted the creation of this post.

The text above will be added to the catalog README with the next release.