In data clustering, many approaches have been proposed such as K-means method and hierarchical method. One of the problems is that the results depend heavily on initial values and criterion to combine clusters. In this investigation, we propose a new method to cluster stream data while avoiding this deficiency. Here we assume there exists aspects of local regression in data. Then we develop our theory to combine clusters using F values by regression analysis as criterion and to adapt to stream data. We examine experiments and show how well the theory works.