In wireless sensor networks, a significant amount of sensor readings sent from the sensors to the data processing point(s) may be lost or corrupted. In this research we propose a power-aware technique, called CARMA (Combining Association Rule Mining with Average window size), to deal with such a problem. In CARMA, to save battery power on sensors, instead of requesting the sensor nodes (MS), the readings of which are missing, to resend their last readings, an estimation of the missing value(s) is performed by using the values available at the sensors relating to the MS through association rule mining. The paper then presents the performance studies comparing CARMA with existing techniques using the real traffic data collected by the Department of Transportation in Austin, Texas.