Based on the so-called partial continuation model with exact finite measurements, a new stochastic assessment procedure is introduced. For every satellite pair, the temporal correlation coefficient is estimated using the original double-differenced (DD) GPS measurements. And then, the Durbin-Watson test is applied to test specific hypothesis on the temporal correlation coefficient. Unless the test is not significant with a certain significant level, a data transformation is required. These transformed measurements are free of time correlations. For purpose of illustration, two static GPS baseline data sets are analyzed in detail. The experimental results demonstrate that the proposed procedure can mitigate effectively the impact of systematic errors on DD GPS measurements.