We have devised a method of analyzing geomagentic array data which effectively filters out variations caused by atmospheric disturbances, their associated ground currents, and local variations in magnetic susceptibility. This method has application to the detection of tectonomagnetic variations. We use vector field data at one station, from which scalar field data at several stations are predicted. Prediction error statistics are then carefully studied, and suspected tectonomagnetic events are described in terms of the probability of their being random fluctuations in normal data. We have applied this method to data from the U.S. Geological Survey magnetometer array in central Califronia, using vector data from Castle Rock Observatory. Prediction errors appear to be Gaussian. The lower limit to the variance appears to be the sensitivity of the vector data at Castle Rock (the hourly means used are reported to the nearest gamma). No believable tectonomagnetic events occur form January 1, 1975, to February 28, 1975.