This paper presents a series of optimization models for Phasor Measurement Unit (PMU) placement with new considerations. Extended from the formulations based on integer linear programming, issues concerning uncertainty propagation are discussed and an improved model is proposed to alleviate the impact. Furthermore, as it is more valuable in practice with incrementally placing PMUs, a multistage installation framework based on dynamic programming is developed. With a novel index introduced, the presented framework is implemented with uncertainty propagation controlled in a certain level for each stage. Unlike other available approaches, when processing the immediate stage in this paper, current and future installations are simultaneously considered rather than just making decisions from fixed PMU locations. Minimum number of PMUs is guaranteed meanwhile with each installation. In addition, a postprocessing optimization for selection from multiple solutions in integer programming models is also presented. The proposed models are tested in IEEE 14-, 30-, 57-and 118-bus systems. With the results analyzed and discussed, it indicates the improved model can limit the uncertainty propagation in a fixed level, and the proposed multistage framework could be employed as a practical strategy for PMU placement planning.