Structural parameters obtained by different methods are compared in Tables I-IV. Since geometries resulting from DH6D and DH5D basis sets are practically identical, only results of RHF and MP2 calculations for DH6D set are reported.
It is well known that bond lengths calculated by the Hartree-Fock method are
usually too short compared to experimental values, while the corresponding
MP2 values 
are much closer to reality. It was also
supported by these calculations and clearly expressed for the polarized
C-O and O-H bonds. This effect was smaller for C-N and N-H bonds and quite
small for C-C and C-H bonds. Bond lengths from DFT bracket
ab initio values on both sides. Bonds involving
hydrogen (C-H, N-H or O-H) which
resulted from LSD calculations are the longest of all, and the effect is
most visible for the O-H bonds and the smallest for the C-H bonds.
The situation is opposite for bonds between the carbon atom and another
1st row element. These bonds calculated with LSD are the shortest of all
methods and the magnitude of this effect follows bond polarity.
Becke-Perdew corrections substantially improve bond lengths. The H-X bonds are now shorter than for the case of uncorrected LSD. The C-X bonds with Becke-Perdew corrections are longer than their LSD counterparts. This brings bond lengths calculated with Becke-Perdew corrections to a much closer agreement with experimental and MP2 values.
Valence angles usually follow the known trend in which the angle between shorter bonds tends to be larger, while angle between overestimated bonds is frequently smaller than it should be. This trend is evident if one compares HF and MP2 results. It is also pronounced for H-X-Y angles, which are usually too small for LSD calculations due to the underestimation of X-Y bond length.
Generally, the structural parameters calculated by different methods are similar. Larger discrepancies are visible for torsional angles in ethanol and protonated ethanol. It is not surprising, however, since the potential energy surface for torsional angles pivoted on the C-O bond of ethanol is very flat, and it is still being debated if the minimum energy of ethanol corresponds to a gauche or trans conformation.