We present results of a statistical analysis of the SFI catalog of peculiar
velocities, a recently completed survey of spiral field galaxies with I-ba
nd Tully-Fisher distances. The velocity field statistic utilized is the vel
ocity correlation function, psi(1)(r), originally introduced by Gorski et a
l. The analysis is performed in redshift space so as to circumvent potentia
l ambiguities connected with inhomogeneous Malmquist bias corrections. The
results from the SFI sample are compared with linear-theory predictions for
a class of cosmological models. We generate a large set of mock samples, e
xtracted from N-body simulations, which are used to assess the reliability
of our analysis and to estimate the associated uncertainties. We assume a c
lass of cold dark matter-like power spectrum models, specified by sigma(8),
the rms fluctuation amplitude within a sphere of 8 h(-1) Mpc radius, and b
y the shape parameter, Gamma. Defining eta(8) = sigma(8) Omega(0)(0.6), we
find that the measured psi(1)(r) implies a degenerate constraint in the (et
a(8),Gamma)-plane, with eta(8) = 0.3 +/- 0.1(Gamma/0.2)(0.5) at the 2 sigma
level for the inverse Tully-Fisher (ITF) calibration presented in this pap
er. We investigate how much this constraint changes as we account for uncer
tainties in the analysis method and uncertainties in the distance indicator
, and we consider alternative ITF calibrations. We find that both changing
the error-weighting scheme and selecting galaxies according to different li
miting line widths has a negligible effect. On the contrary, the model cons
traints are quite sensitive to the ITF calibration. The other ITF calibrati
ons, by Giovanelli et al. and da Costa et al, both yield, for Gamma = 0.2,
a best-fit value of eta(8) similar or equal to 0.6.