A description of the fracture geometry pattern is a primary requirement to assess the hydraulic and mechanical behaviors of jointed rock masses. In order to enhance the reliability of fracture data, this paper proposes some analytic techniques based on statistical approaches to characterize and analyze the variability of fracture pattern. Various distributed forms including the goodness of fit tests are used to model fracture data (such as orientation, spacing, trace length and interconnectivity index) distribution. An extensive fracture database was developed for andesite rock at Lan‐Yu, Taiwan, to support work on the fracture hydrology characteristics of jointed rock masses. Results based on cluster analysis show that fracture orientation data can be delineated by four sets at the Lan‐Yu site. The best fit form for orientation distribution of each set was not found, but the Fisher distribution is suggested due to its simple form. The best fit distributions of spacings and trace lengths can respond to exponential and log‐normal, respectively. The results also indicate that one of the four fracture sets has a larger variation than the others according to spacing analysis. The same trend occurs in the analysis of fracture trace length. All have mean interconnectivity values larger than unity, indicating that the four fracture sets provide hydraulic communication within the fracture network.