Diffusive transport through porous materials is to a large extent determined by the microstructure of the material. To design materials with controlled transport properties, it is hence important to connect properties of the pore geometry to diffusive transport rates. Different kinds of microstructures from a stochastic model are generated and multiplicative regression is used to find relationships between geometric measures of the microstructures and numerically simulated diffusive transport. The main results are that the geodesic tortuosity explains a large part of the transport variation, and that the standard deviations we introduce further improves prediction. It was found that it is best to compute the tortuosity using the whole pore space, instead of using only the inlet, as is commonly done. The effects of calculating the measures using small samples of the pore structure were investigated, and a method for minimizing errors resulting from boundary effects was proposed.