Abstract: In road engineering, the pavement mainly bears the action of vehicle loads, and the pavement structure should have enough strength, stiffness, durability and other properties. Therefore, the performance of pavement structure materials is an important guarantee that affects the quality of pavement service. At present, the inorganic bonding stable materials are often used in highway engineering, and their strength is often measured by the unconfined compressive strength tests. The 3 times standard deviation method (3S method) is recommended for the treatment of outliers. Since the unconfined compressive strength test sample size is small, the 3S method cannot eliminate some specific data from the test data. In order to solve this problem, the new method of t distribution confidence interval is proposed, and it is analyzed theoretically. By comparing other treatment handling methods for outliers, the t distribution confidence interval analysis method in the treatment of the unconfined compressive strength test data is confirmed to be feasible, and the proposed method guarantees the possible range of size and ensures the overall index to fall within the scope of the probability.