Eddy current (EC) methods of nondestructive evaluation are used extensively in steam generator tube inspection in nuclear power plants. With advances in technology, new sensor designs are being proposed for improving data acquisition and defect detection. Limitations associated with manual analysis of the large volume of EC data have led to the development of automated data analysis. In this paper, we present statistical models for automated analysis of the complex array probe data. The EC data is modeled as independent random ariables with identical distribution and the statistical parameters of the two class problem are estimated using Maximum Likelihood (ML) method. Defect identification in the noisy EC data is assessed using minimum error rate (MER) detection rule. Receiver operating characteristics (ROC) generated for in-service steam generator tubes are used to assess the model performance.