Most radio positioning methods are based on the measurements of distance between different wireless nodes. Owing to the existence of non-line-of-sight (NLOS) radio propagation, unfortunately, not all the measured distances are reliable. One way to tackle the problem of positioning is therefore to take two-steps: (i) identifying the NLOS measurements; (ii) smart signal processing of the mixed LOS and NLOS measurements. This paper is focused on the second issue. Under the assumption that the NLOS measurements have been identified, we first propose a simple method to suppress the effect of the NLOS error. Simulation results demonstrate that the proposed method achieves similar or better accuracy than several other known methods and the computational complexity is reduced considerably. We also present an optimal location estimator under the assumption of Gaussian distributed measurement noise and Rayleigh distributed NLOS error. Although it is difficult to achieve the optimal performance in practice due to modeling uncertainties, the optimal estimator provides a performance benchmark.