Abstract: The accuracies of different satellite observations are different since the effects of the atmosphere delay,multipath and diffraction are different;so the stochastic model is very important in precise dynamic GPS positioning.The satellite elevation angle,C/No(carrier-to-noise-power density ratio),signal strength are the important indices of the observation quality.The stochastic models based on these indices can mitigate the errors such as diffraction;but each model is suitable to process some kinds of errors.In order to use these models correctly,some computing tests using real GPS data are carried out.The results show that the model based on signal strength as power as the model based on C/No.It means that the signal strength can substitute C/No as the indea in SIGMA stochastic model when there is no C/No output in observation data.The comparsion of these models indicates that models based on signal strength and C/No can mitigate diffraction errors more,while the model based on satellite elevation angle can mitigate more residual tropospheric delay errors.