Many uncertainties of power system parameters such as loads， generations and network parameters will affect power system small signal stability analysis. A credibility theory based method is proposed to consider these uncertainties for small signal stability assessment. The method can be used to calculate the expectations and variances of eigenvalues of differential-algebraic equations. It can also be used to calculate the small signal stability indices based on credibility theory. The results provide useful information for system operation and planning decision-making. The credibility theory method avoids hypothesis of the uncertain parameters distributed function's type and the function parameters' statistics， and does not require independence of uncertain parameters. The membership function of uncertain parameters is required， but it has no influence on the result. Compared with deterministic analysis method， two point method and Monte Carlo simulation method， case studies verify its feasibility and applicability for power system small signal stability assessment.