为了获得组合GPS／GLONASS精密定位结果，该文从理论和数值实例两方面分析了研究组合GPS／GLONASS观测值随机模型的重要性，提出了两种利用观测值的误差残差估计随机模型的方法，即验后估计法和方差－协方差迭代法。理论和数值结果表明，这两种随机模型估计方法与采用经验随机模型相比，可以提高整周模糊度解算的可靠性和定位精度，所提出的观测值随机模型估计方法理论上更严格，实践上可行，并建议采用方差－协方差迭代法估计组合GPS／GLONASS精密定位的观测值随机模型。

The processing of GPS/GLONASS data requires the specification of an observation model, which consisted of a functional and a stochastic model. The GPS/GLONASS functional model is already well known, but the GPS/GLONASS stochastic model is not. The stochastic model depends on not only the types of the observation and the receiver used, but also the environmental effects. To obtain the precisely combined GPS and GLONASS positioning, the importance of stochastic model of combined GPS and GLONASS observed value is analyzed. The posterior estimation method and iteration of variance and covariance method to estimate stochastic model using residuals of observable errors are proposed. The theoretical and numerical results show that the proposed method which can improve the reliability and precision of ambiguity resolution, is more rigorous in theory and more powerful in practice comparing with the stochastic model, Using iteration of variance and covariance method to estimate stochastic model of combined GPS and GLONASS observables is suggested.