The paper examines the accuracy of various model reduction techniques via modal analysis of a truss structure. Besides the widely used Craig-Bampton (CB) method in commercial software, three additional methods - modal truncation augmentation (MTA), system equivalent reduction expansion process (SEREP), and the Rubin method - are briefly introduced. Subsequently, the results provided by these methods are investigated through the comparison of natural frequencies, mode shapes, and frequency response functions. The study reveals that generally, the Rubin method provides the most accurate results, except in higher frequency ranges where its error suddenly spikes. With the MTA method, accuracy can be improved at low frequencies, but the error gradually increases. In the case of SEREP, the error remains approximately constant, resulting in larger errors at low frequencies and smaller errors at high frequencies compared to other methods.