A mercury intrusion experiment is one of the most important methods to investigate pore structure of core samples by measuring petrophysical parameters, such as the sorting coefficient and mean pore radius, etc. These petrophysical parameters describe the pore structure, which has a strong relationship with porosity and permeability. However there is no universal mathematical relationship between them so far. With a better understanding of such a relationship, reservoir simulators can be built more accurately. Since the oilfields' development decades ago, mercury intrusion data have been collected from a variety of geological settings and the different geological characteristics often have substantial impacts on the relationship between the petrophysical parameters. Therefore, we aim to develop a statistical model that not only captures this relationship but also accounts for different geological backgrounds. Data containing more than thirty development blocks were collected and classified. Based on these data, a mixed model was developed to describe this hierarchical behavior. An overall connection between the petrophysical parameters is treated as the fixed effect at a higher level, whereas some small variations due to the geological background are accounted for as the random effect at a lower level. Although all petrophysical parameters are considered in the model as either the fixed or random effect at the beginning, each one is tested for its significance. If results show that the random effect is insignificant, then this parameter has no noticeable difference across different geologic settings. If the fixed effect is insignificant, then this parameter has no critical impact on the relationship. Because a statistical model is built, the results are shown probabilistically. Essential visualizations are presented to help understand and interpret the results. Variable selection and hypothesis testing are performed to develop the most desirable model. A final best model is achieved to explain the relationship based on different geological behaviors, and this model can be used for prediction in different geological settings. With the studied relationship and advanced model, reservoir simulators can be greatly refined and improved in the future.