This paper first gives an analysis of Data Aggregation and Data Compression based on the energy consumption of sensor nodes, after which an approach is proposed to construct an aggregation tree in the case of non-perfect aggregation since GIT considers only the case of perfect aggregation and it does not work well if the aggregation is non-perfect. An assessment scheme that can get the information of hops from the aggregation point to the sink and the hops from the aggregation point to the source node is used to construct such an aggregation tree. Moreover, the energy consumption of the aggregation is also considered. This scheme can be used when perfect aggregation cannot be performed. In this paper, an approach to reduce the cost of reinforcement is also proposed, in which the reinforcement work is done by the source nodes themselves, not by the sink node. Simulation result shows that this approach can save more energy than GIT when the aggregation ratio is small. This result also provides a theoretical limit of aggregation to tell when GIT will lose its superiority and thus gives us a direction to choose among the aggregation algorisms. Another result shows that the further the sources are away from the sink, the less reinforcement messages are needed.