This paper describes a procedure for structural health assessment based on one-wayanalysis of variance (ANOVA) together with Tukey's multiple comparison test, todetermine whether the results are statistically significant. The feature indices are obtainedfrom electromechanical impedance measurements using piezoceramic sensor/actuatorpatches bonded to the structure. Compared to the classical approach based on a simplechange of the observed signals, using for example root mean square responses, thedecision procedure in this paper involves a rigorous statistical test. Experimental testswere carried out on an aeronautical panel in the laboratory to validate the approach. Inorder to include uncontrolled variability in the dynamic responses, the measurementswere taken over several days in different environmental conditions using all eight sensor/actuator patches. The damage was simulated by controlling the tightness and looseness ofthe bolts and was correctly diagnosed. The paper discusses the strengths and weakness ofthe approach in light of the experimental results.