In a previous work, we introduced a three-parameter scaling solution that models the long-term recovery of dry gas from a hydrofractured horizontal well far from other wells and the boundaries of a shale reservoir with negligible sorption. Here, we extend this theory to account for the contribution of sorbed gas and apply the extended theory to the production histories of 8,942 dry-gas wells in the Marcellus Shale. Our approach is to integrate unstructured big data and physics-based modeling. We consider three adsorption cases that correspond to the minimum, median, and maximum of a set of measured Langmuir isotherms. We obtain data-driven, independent estimates of unstimulated shale permeability, spacing between hydrofractures, well-drainage area, optimal spacing between infill wells, and incremental gas recovery over a typical well life. All these estimates decrease to varying extents with increasing sorption. We find that the average well with median adsorption has a permeability of 250 nd, fracture spacing of 16 m, 30-year drainage length of 79 m, and a 30-year incremental recovery of 67%.