We have investigated the value of isotropic seismic converted-wave (i.e., PS) data for reservoir parameter estimation using stochastic approaches based on a floating-grain rock-physics model. We first performed statistical analysis on a simple two-layer model built on actual borehole logs and compared the relative value of PS data versus amplitude-variation-with-offset (AVO) gradient data for estimating the floating-grain fraction. We found that PS data were significantly more informative than AVO gradient data in terms of likelihood functions, and the combination of PS and AVO gradient data together with PP data provided the maximal value for the reservoir parameter estimation. To evaluate the value of PS data under complex situations, we developed a hierarchical Bayesian model to combine seismic PP and PS data and their associated time registration. We extended a model-based Bayesian method developed previously for inverting seismic PP data only, by including PS responses and time registration as additional data and PS traveltime and reflectivity as additional variables. We applied the method to a synthetic six-layer model that closely mimics real field scenarios. We found that PS data provided more information than AVO gradient data for estimating the floating-grain fraction, porosity, net-to-gross, and layer thicknesses when their corresponding priors were weak.