||The objective of this paper is to develop and evaluate an inversion algorithm for estimating the parameters in a geoacoustic model for a marine sediment layer. The geoacoustic model employs a generalized-exponential function for sound speed profile, and an inverse-square function for density distribution. Based upon plane-wave reflection from a nonuniform sediment layer overlying an elastic seafloor, an inversion procedure is established and numerically implemented for estimating the parameters using synthetic noisecontaminated data. The sensitivity of each model parameter is first studied, and then three highly sensitive parameters in the geoacoustic model and one statistical parameter are chosen for inversion analysis. The resulting sound speed profiles from the inversion are analyzed by a probabilistic approach, which is quantified by the posterior probability density for the uncertainties of the estimated parameters. The parameter uncertainties referenced to 1-D and 2-D marginal posterior probability densities are investigated, followed by the statistical estimation for the sound speed profile in terms of a 95% credibility interval. We demonstrate the effects of, the signal- to-noise ratio (SNR), the dimension of the data vector, and the region in which the data are sampled, on the statistical estimation of the sound speed profile, and offer physical interpretations about the statistical variations attributable to these effects. This analysis provides a basis for estimating the acoustic properties of a continuously-stratified layer using inversion approach, and is particularly useful for a medium with properties describable by analytical functions.