We address two of the canonical problems in sensor networks: data integrity and computational sensing. Due to the large scale and distributed nature of sensor networks, their heterogeneous node structure, cost and power constraints, operation in unpredictable and unconditioned (and often harsh) environmental surroundings and inherent unreliability of sensors sensor networks often collect data with errors, faults and missing samples. We have developed a generic approach for both tasks that has three phase: (i) statistical modeling, (ii) prediction, and (iii) fusion and analysis.