Wireless sensor networks are used to monitor a given environment, such as indoor heating conditions or the micro-climate of glaciers. They offer a low-cost solution that provides a high data density. Usually the user of such a sensor network has a good idea of how, knowing the environment, the sensed values should behave. This idea can be expressed as a data model. Such models can be used to detect anomalies, compress data, or combine data from many inexpensive sensors to increase the quality of the measurements. This paper presents a framework to process arbitrary sensor-network data models. The framework can then be used to distribute the model processing into the wireless sensor network. Quality of information criteria are used to determine the performance of the models. A prototype of the framework is presented together with a comparison of two existing stochastic data model approaches for wireless sensor networks.