Abstract: Based on wind tunnel tests of a rigid model of a railway station building, high-order statistics of long-duration pressure processes were quantified in terms of skewness and kurtosis. It was found that the estimation of kurtosis is sensitive to the duration of wind pressure time-history data. Depending on whether the kurtosis is larger or smaller than three, non-Gaussian random processes can be classified as softening or hardening processes, respectively. Hardening load processes with kurtosis values smaller than three were observed in samples of wind pressure acting on the long-span roof. Non-Gaussian peak factors of wind pressure, particularly for those observed hardening pressure process, were calculated by employing various state-of-the-art methods and compared to the direct statistical analysis of the measured long-duration wind pressure data. The results suggested that kurtosis should not be used in the calculation of non-Gaussian peak factors. The recently proposed translated peak process (TPP) method is able to estimate non-Gaussian peak factors for both hardening and soft pressure processes.