We present an analysis of some of the properties of the density field reali
zed in numerical simulations for power-law initial power spectra in the cas
e of a critical density universe. We study the non-linear regime, which is
the most difficult to handle analytically, and we compare our numerical res
ults with the predictions of a specific hierarchical clustering scaling mod
el that have been made recently, focusing specifically on its much wider ra
nge of applicability, which is one of its main advantages over the standard
Press-Schechter approximation. We first check that the two-point correlati
on functions, measured from both counts-in-cells and neighbour counts, agre
e with the known analytically exact scaling requirement (i.e., depend only
on sigma(2)), and we also find the stable-clustering hypothesis to hold. Ne
xt, we show that the statistics of the counts-in-cells obey the scaling law
predicted by the above scaling model.
Then we turn to mass functions of overdense and underdense regions, which w
e obtain numerically from 'spherical overdensity' and 'friends-of-friends'
algorithms. We first consider the mass function of 'just-collapsed' objects
defined by a density threshold Delta = 177, and we note, as was found by p
revious studies, that the usual Press-Schechter prescription agrees reasona
bly well with the simulations (although there are some discrepancies). On t
he other hand, the numerical results are also consistent with the predictio
ns of the scaling model. Next, we consider more general mass functions (nee
ded to describe for instance galaxies or Lyman-alpha absorbers) defined by
different density thresholds, which can even be negative. The scaling model
is especially suited to account for such cases, which are out of reach of
the Press-Schechter approach, and it still shows reasonably good agreement
with the numerical results. Finally, we show that mass functions defined by
a condition on the radius of the objects also satisfy the theoretical scal
ing predictions.
Thus we find that the scaling model provides a reasonable description of th
e density field in the highly non-linear regime, for the cosmologies we hav
e considered, for both the counts-in-cells statistics and the mass function
s. The advantages of this approach are that it clarifies the links between
several statistical tools and it allows one to study many different classes
of objects, for any density threshold, provided one is in the fully non-li
near regime.